diff --git a/litellm/litellm_core_utils/reasoning_effort_utils.py b/litellm/litellm_core_utils/reasoning_effort_utils.py
new file mode 100644
index 000000000000..5987392d070c
--- /dev/null
+++ b/litellm/litellm_core_utils/reasoning_effort_utils.py
@@ -0,0 +1,26 @@
+from typing import Literal
+
+from litellm.constants import (
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+)
+
+OpenAIStyleReasoningEffort = Literal["minimal", "low", "medium", "high"]
+
+
+def reasoning_effort_from_thinking_budget(
+    budget_tokens: int,
+) -> OpenAIStyleReasoningEffort:
+    """Bucket an Anthropic ``thinking.budget_tokens`` into an OpenAI-style
+    ``reasoning_effort`` using the shared ``DEFAULT_REASONING_EFFORT_*_THINKING_BUDGET``
+    thresholds, so every backend that translates a budget into an effort label
+    reads the same numbers.
+    """
+    if budget_tokens >= DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET:
+        return "high"
+    if budget_tokens >= DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET:
+        return "medium"
+    if budget_tokens >= DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET:
+        return "low"
+    return "minimal"
diff --git a/litellm/llms/anthropic/chat/transformation.py b/litellm/llms/anthropic/chat/transformation.py
index 2b2264e6da97..b223aef1415e 100644
--- a/litellm/llms/anthropic/chat/transformation.py
+++ b/litellm/llms/anthropic/chat/transformation.py
@@ -229,10 +229,7 @@ def _build_anthropic_tool_name_maps(
     "Sonnet 4.6+, and Mythos Preview."
 )
 
-DROP_UNSUPPORTED_SPEED_WARNING = (
-    "Dropping unsupported `speed` for model=%s "
-    "(drop_params=True). Fast mode is only supported on select Opus models."
-)
+DROP_UNSUPPORTED_SPEED_WARNING = "Dropping unsupported `speed` for model=%s (drop_params=True). Fast mode is only supported on select Opus models."
 
 
 class AnthropicConfig(AnthropicModelInfo, BaseConfig):
@@ -352,8 +349,7 @@ def _supports_effort_level(model: str, level: str) -> bool:
     def _validate_effort_for_model(model: str, effort: Optional[str]) -> Optional[str]:
         """Return ``None`` if ``effort`` is allowed on ``model``, else an error message."""
         if effort == "max" and not (
-            AnthropicConfig._is_claude_4_6_model(model)
-            or AnthropicConfig._is_claude_4_7_model(model)
+            AnthropicConfig._is_adaptive_thinking_model(model)
             or AnthropicConfig._supports_effort_level(model, "max")
         ):
             return f"effort='max' is not supported by this model. Got model: {model}"
@@ -471,8 +467,7 @@ def get_supported_openai_params(self, model: str):
 
         if (
             "claude-3-7-sonnet" in model
-            or AnthropicConfig._is_claude_4_6_model(model)
-            or AnthropicConfig._is_claude_4_7_model(model)
+            or AnthropicConfig._is_adaptive_thinking_model(model)
             or supports_reasoning(
                 model=model,
                 custom_llm_provider=self.custom_llm_provider,
@@ -2093,8 +2088,7 @@ def _apply_output_config(
         if effort is not None and effort not in valid_efforts:
             raise litellm.exceptions.BadRequestError(
                 message=(
-                    f"Invalid effort value: {effort!r}. Must be one of: "
-                    f"'high', 'medium', 'low', 'xhigh', 'max'"
+                    f"Invalid effort value: {effort!r}. Must be one of: 'high', 'medium', 'low', 'xhigh', 'max'"
                 ),
                 model=model,
                 llm_provider=self.custom_llm_provider or "anthropic",
diff --git a/litellm/llms/anthropic/common_utils.py b/litellm/llms/anthropic/common_utils.py
index eaa762204f22..06b598c525bd 100644
--- a/litellm/llms/anthropic/common_utils.py
+++ b/litellm/llms/anthropic/common_utils.py
@@ -26,6 +26,23 @@
 )
 from litellm.types.llms.openai import AllMessageValues
 
+_BEDROCK_VERSION_SUFFIX_RE = re.compile(r"-v\d+(?::\d+)?$")
+_INFERENCE_PROFILE_MINOR_RE = re.compile(r":\d+$")
+_DATED_RELEASE_SUFFIX_RE = re.compile(r"-\d{8}$")
+_DOTTED_VERSION_RE = re.compile(r"(\d)\.(\d)")
+
+
+def _strip_bedrock_id_suffixes(model: str) -> str:
+    """Reduce a full Bedrock model id to its base cost-map key by rewriting a
+    dotted family version then peeling a trailing ``-vN:rev`` and ``-YYYYMMDD``
+    in that order, so the real ``-<date>-v1:0`` shape (e.g.
+    ``us.anthropic.claude-sonnet-4-6-20251101-v1:0``) resolves rather than only
+    the date or version in isolation."""
+    return _DATED_RELEASE_SUFFIX_RE.sub(
+        "",
+        _BEDROCK_VERSION_SUFFIX_RE.sub("", _DOTTED_VERSION_RE.sub(r"\1-\2", model)),
+    )
+
 
 def is_anthropic_oauth_key(value: Optional[str]) -> bool:
     """Check if a value contains an Anthropic OAuth token (sk-ant-oat*)."""
@@ -243,38 +260,6 @@ def is_input_examples_used(self, tools: Optional[List]) -> bool:
 
         return False
 
-    @staticmethod
-    def _is_claude_4_6_model(model: str) -> bool:
-        """Check if the model is a Claude 4.6 model (Opus 4.6 or Sonnet 4.6)."""
-        model_lower = model.lower()
-        return any(
-            v in model_lower
-            for v in (
-                "opus-4-6",
-                "opus_4_6",
-                "opus-4.6",
-                "opus_4.6",
-                "sonnet-4-6",
-                "sonnet_4_6",
-                "sonnet-4.6",
-                "sonnet_4.6",
-            )
-        )
-
-    @staticmethod
-    def _is_claude_4_7_model(model: str) -> bool:
-        """Check if the model is a Claude 4.7 model (Opus 4.7)."""
-        model_lower = model.lower()
-        return any(
-            v in model_lower
-            for v in (
-                "opus-4-7",
-                "opus_4_7",
-                "opus-4.7",
-                "opus_4.7",
-            )
-        )
-
     @staticmethod
     def _supports_sampling_params(model: str) -> bool:
         """Claude 4.7+ (Opus 4.7/4.8, Fable 5) removed sampling params: the API
@@ -336,27 +321,42 @@ def _apply_sampling_param(
 
     @staticmethod
     def _model_map_lookup_candidates(model: str) -> List[str]:
-        """Model-map keys to try for ``model``, stripping bedrock/vertex
-        prefixes so a provider-routed Claude still resolves to its entry."""
-        candidates = [model]
-        for prefix in (
+        """Model-map keys to try for ``model``: the id itself, the same id with a
+        bedrock/vertex routing prefix removed, the Bedrock base model, and each of
+        those normalized by stripping a Bedrock version suffix (``-v1:0`` fully or
+        just the ``:0`` inference-profile minor), stripping a dated-release suffix
+        (``-20260205``), or rewriting a dotted family version to hyphens
+        (``4.6`` -> ``4-6``). Lets any reasonable alias (e.g.
+        ``bedrock/invoke/global.anthropic.claude-opus-4-7-v1:0``,
+        ``claude-sonnet-4-6-20260219`` or ``claude-sonnet-4.6``) resolve to its base
+        cost-map entry so the capability flag on that entry stays authoritative."""
+        prefixes = (
             "bedrock/converse/",
             "bedrock/invoke/",
             "bedrock/",
             "vertex_ai/",
-        ):
-            if model.startswith(prefix):
-                candidates.append(model[len(prefix) :])
+        )
+        deprefixed = tuple(model[len(p) :] for p in prefixes if model.startswith(p))
         try:
             from litellm.llms.bedrock.common_utils import BedrockModelInfo
 
             base = BedrockModelInfo.get_base_model(model)
-            if base:
-                candidates.append(base)
-                candidates.append(f"bedrock/{base}")
         except Exception:
-            pass
-        return candidates
+            base = None
+        bedrock_base = (base, f"bedrock/{base}") if base else ()
+        primary = (model, *deprefixed, *bedrock_base)
+        normalized = tuple(
+            stripped
+            for cand in primary
+            for stripped in (
+                _BEDROCK_VERSION_SUFFIX_RE.sub("", cand),
+                _INFERENCE_PROFILE_MINOR_RE.sub("", cand),
+                _DATED_RELEASE_SUFFIX_RE.sub("", cand),
+                _DOTTED_VERSION_RE.sub(r"\1-\2", cand),
+                _strip_bedrock_id_suffixes(cand),
+            )
+        )
+        return list(dict.fromkeys((*primary, *normalized)))
 
     @staticmethod
     def _get_model_capability(model: str, key: str) -> Optional[bool]:
@@ -403,19 +403,16 @@ def _supports_model_capability(model: str, key: str) -> bool:
 
     @staticmethod
     def _is_adaptive_thinking_model(model: str) -> bool:
-        """Claude 4.6+ models use adaptive thinking with ``output_config.effort``.
+        """Whether ``model`` uses adaptive thinking (``output_config.effort``).
 
-        Driven by the ``supports_adaptive_thinking`` flag in the model map; the
-        4.6/4.7 name checks remain only as a fallback for provider-routed ids
-        whose map entries predate the flag.
+        Sourced solely from the model cost map's ``supports_adaptive_thinking`` flag,
+        resolved through provider prefixes. A model that resolves to no mapped entry
+        (an unmapped alias or a future release not yet in the map) is treated as
+        non-adaptive until a ``fallback_generalizations`` rule covers it.
         """
-        if AnthropicModelInfo._supports_model_capability(
+        return AnthropicModelInfo._supports_model_capability(
             model, "supports_adaptive_thinking"
-        ):
-            return True
-        return AnthropicModelInfo._is_claude_4_6_model(
-            model
-        ) or AnthropicModelInfo._is_claude_4_7_model(model)
+        )
 
     def is_effort_used(
         self, optional_params: Optional[dict], model: Optional[str] = None
diff --git a/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py b/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
index 46564a565cec..dd8d0d8d50ec 100644
--- a/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
+++ b/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
@@ -76,6 +76,9 @@ def create_tool_name_mapping(
 from litellm.litellm_core_utils.prompt_templates.factory import (
     THOUGHT_SIGNATURE_SEPARATOR,
 )
+from litellm.litellm_core_utils.reasoning_effort_utils import (
+    reasoning_effort_from_thinking_budget,
+)
 from litellm.llms.anthropic.common_utils import normalize_anthropic_tool_use_id
 from litellm.llms.anthropic.experimental_pass_through.context_management import (
     PolyfillResult,
@@ -716,11 +719,8 @@ def translate_anthropic_thinking_to_reasoning_effort(
         Anthropic thinking format: {'type': 'enabled'|'disabled', 'budget_tokens': int}
         OpenAI reasoning_effort: 'none' | 'minimal' | 'low' | 'medium' | 'high' | 'xhigh' | 'default'
 
-        Mapping:
-        - budget_tokens >= 10000 -> 'high'
-        - budget_tokens >= 5000  -> 'medium'
-        - budget_tokens >= 2000  -> 'low'
-        - budget_tokens < 2000   -> 'minimal'
+        ``budget_tokens`` is bucketed via the shared
+        ``reasoning_effort_from_thinking_budget`` thresholds.
         """
         if not isinstance(thinking, dict):
             return None
@@ -730,15 +730,9 @@ def translate_anthropic_thinking_to_reasoning_effort(
         if thinking_type == "disabled":
             return None
         elif thinking_type == "enabled":
-            budget_tokens = thinking.get("budget_tokens", 0)
-            if budget_tokens >= 10000:
-                return "high"
-            elif budget_tokens >= 5000:
-                return "medium"
-            elif budget_tokens >= 2000:
-                return "low"
-            else:
-                return "minimal"
+            return reasoning_effort_from_thinking_budget(
+                thinking.get("budget_tokens", 0)
+            )
         elif thinking_type == "adaptive":
             # Adaptive thinking: effort is controlled by output_config.effort,
             # not budget_tokens. Return a default; caller should override with
diff --git a/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py b/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py
index 07e8270b496c..fbe49ae41366 100644
--- a/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py
+++ b/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py
@@ -2,6 +2,11 @@
 
 import httpx
 
+from litellm.constants import (
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET,
+)
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
 from litellm.litellm_core_utils.litellm_logging import verbose_logger
 from litellm.llms.base_llm.anthropic_messages.transformation import (
@@ -248,11 +253,13 @@ def _translate_legacy_thinking_for_adaptive_model(
             return
 
         budget = int(thinking.get("budget_tokens") or 0)
-        if budget >= 24000 and AnthropicConfig._supports_effort_level(model, "xhigh"):
+        if budget >= DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET and (
+            AnthropicConfig._supports_effort_level(model, "xhigh")
+        ):
             effort = "xhigh"
-        elif budget >= 10000:
+        elif budget >= DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET:
             effort = "high"
-        elif budget >= 5000:
+        elif budget >= DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET:
             effort = "medium"
         else:
             effort = "low"
diff --git a/litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py b/litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py
index 4fb1ddf5c46d..95ab9b3a0fe9 100644
--- a/litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py
+++ b/litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py
@@ -8,6 +8,9 @@
 import json
 from typing import Any, Dict, List, Optional, Union, cast
 
+from litellm.litellm_core_utils.reasoning_effort_utils import (
+    reasoning_effort_from_thinking_budget,
+)
 from litellm.llms.anthropic.experimental_pass_through.utils import (
     is_reasoning_auto_summary_enabled,
 )
@@ -257,11 +260,10 @@ def translate_thinking_to_reasoning(
         """
         Convert Anthropic thinking param to Responses API reasoning param.
 
-        thinking.budget_tokens maps to reasoning effort:
-          >= 10000 -> high, >= 5000 -> medium, >= 2000 -> low, < 2000 -> minimal
-
-        For adaptive thinking, uses output_config.effort if available,
-        otherwise defaults to medium.
+        ``thinking.budget_tokens`` is bucketed via the shared
+        ``reasoning_effort_from_thinking_budget`` thresholds. For adaptive
+        thinking, uses ``output_config.effort`` if available, otherwise defaults
+        to medium.
         """
         if not isinstance(thinking, dict):
             return None
@@ -274,15 +276,9 @@ def translate_thinking_to_reasoning(
             if isinstance(output_config, dict) and output_config.get("effort"):
                 effort = output_config["effort"]
         elif thinking_type == "enabled":
-            budget = thinking.get("budget_tokens", 0)
-            if budget >= 10000:
-                effort = "high"
-            elif budget >= 5000:
-                effort = "medium"
-            elif budget >= 2000:
-                effort = "low"
-            else:
-                effort = "minimal"
+            effort = reasoning_effort_from_thinking_budget(
+                thinking.get("budget_tokens", 0)
+            )
         else:
             return None
 
diff --git a/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py b/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py
index 42c3bd517a91..3ce426edba76 100644
--- a/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py
+++ b/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py
@@ -14,7 +14,12 @@
 
 import litellm
 from litellm.anthropic_beta_headers_manager import filter_and_transform_beta_headers
-from litellm.constants import BEDROCK_MIN_THINKING_BUDGET_TOKENS
+from litellm.constants import (
+    BEDROCK_MIN_THINKING_BUDGET_TOKENS,
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET,
+)
 from litellm.litellm_core_utils.litellm_logging import verbose_logger
 from litellm.llms.anthropic.chat.transformation import (
     DROP_UNSUPPORTED_OUTPUT_CONFIG_WARNING,
@@ -80,6 +85,51 @@ def __init__(self, **kwargs):
         BaseAnthropicMessagesConfig.__init__(self, **kwargs)
         AmazonInvokeConfig.__init__(self, **kwargs)
 
+    @staticmethod
+    def _as_system_content_blocks(value: Any) -> list[Any]:
+        if value is None:
+            return []
+        if isinstance(value, list):
+            return list(value)
+        if isinstance(value, str):
+            return [{"type": "text", "text": value}]
+        return [value]
+
+    def _normalize_system_role_messages_for_bedrock(
+        self, anthropic_messages_request: dict
+    ) -> None:
+        """Bedrock Invoke rejects ``role: "system"`` entries inside ``messages`` on
+        some Claude aliases; Anthropic Messages carries that content in the
+        top-level ``system`` field. Move any such entries into ``system`` before
+        the Invoke request is built."""
+        messages = anthropic_messages_request.get("messages")
+        if not isinstance(messages, list):
+            return
+        system_role_messages = [
+            m for m in messages if isinstance(m, dict) and m.get("role") == "system"
+        ]
+        if not system_role_messages:
+            return
+
+        anthropic_messages_request["messages"] = [
+            m
+            for m in messages
+            if not (isinstance(m, dict) and m.get("role") == "system")
+        ]
+        system_content = [
+            block
+            for source in (
+                anthropic_messages_request.get("system"),
+                *(m.get("content") for m in system_role_messages),
+            )
+            for block in self._as_system_content_blocks(source)
+        ]
+        filtered_system = self._filter_billing_headers_from_system(system_content)
+        if filtered_system:
+            anthropic_messages_request["system"] = filtered_system
+        else:
+            anthropic_messages_request.pop("system", None)
+
     def validate_anthropic_messages_environment(
         self,
         headers: dict,
@@ -195,8 +245,9 @@ def _supports_extended_thinking_on_bedrock(self, model: str) -> bool:
         """
         Check if the model supports extended thinking beta headers on Bedrock.
 
-        On 3rd-party platforms (e.g., Amazon Bedrock), extended thinking is only
-        supported on: Claude Opus 4.5, Claude Opus 4.1, Opus 4, or Sonnet 4.
+        On 3rd-party platforms (e.g., Amazon Bedrock), extended thinking is supported
+        on the adaptive-thinking models (sourced from the cost map) plus the legacy
+        non-adaptive set: Claude Opus 4.5, Claude Opus 4.1, Opus 4, or Sonnet 4.
 
         Ref: https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking
 
@@ -206,10 +257,11 @@ def _supports_extended_thinking_on_bedrock(self, model: str) -> bool:
         Returns:
             True if the model supports extended thinking on Bedrock
         """
-        model_lower = model.lower()
+        if AnthropicModelInfo._is_adaptive_thinking_model(model):
+            return True
 
-        # Supported models on Bedrock for extended thinking
-        supported_patterns = [
+        model_lower = model.lower()
+        non_adaptive_patterns = [
             "opus-4.5",
             "opus_4.5",
             "opus-4-5",
@@ -222,21 +274,9 @@ def _supports_extended_thinking_on_bedrock(self, model: str) -> bool:
             "opus_4",  # Opus 4
             "sonnet-4",
             "sonnet_4",  # Sonnet 4
-            "sonnet-4.6",
-            "sonnet_4.6",
-            "sonnet-4-6",
-            "sonnet_4_6",
-            "opus-4.6",
-            "opus_4.6",
-            "opus-4-6",
-            "opus_4_6",
-            "opus-4.7",
-            "opus_4.7",
-            "opus-4-7",
-            "opus_4_7",
         ]
 
-        return any(pattern in model_lower for pattern in supported_patterns)
+        return any(pattern in model_lower for pattern in non_adaptive_patterns)
 
     def _ensure_thinking_for_clear_thinking_context_management(
         self,
@@ -270,14 +310,27 @@ def _ensure_thinking_for_clear_thinking_context_management(
         if not self._supports_extended_thinking_on_bedrock(model):
             return False
 
+        is_adaptive_thinking_model = AnthropicModelInfo._is_adaptive_thinking_model(
+            model
+        )
+
         thinking = anthropic_messages_request.get("thinking")
         if isinstance(thinking, dict):
             t = thinking.get("type")
-            if t in ("enabled", "adaptive"):
+            if t == "adaptive":
+                return False
+            if t == "enabled" and not is_adaptive_thinking_model:
                 return False
-            # ``disabled`` or unknown — replace with enabled so clear_thinking is valid
+            if t == "enabled":
+                budget_tokens = self._resolve_clear_thinking_budget_tokens(
+                    thinking.get("budget_tokens")
+                )
+                self._inject_adaptive_thinking_for_clear_thinking(
+                    anthropic_messages_request, budget_tokens, model
+                )
+                return True
             verbose_logger.debug(
-                "Bedrock clear_thinking_20251015: replacing thinking=%s with minimal enabled thinking",
+                "Bedrock clear_thinking_20251015: replacing thinking=%s with minimal thinking config",
                 thinking,
             )
 
@@ -292,6 +345,12 @@ def _ensure_thinking_for_clear_thinking_context_management(
             )
             return False
 
+        if is_adaptive_thinking_model:
+            self._inject_adaptive_thinking_for_clear_thinking(
+                anthropic_messages_request, budget, model
+            )
+            return True
+
         anthropic_messages_request["thinking"] = {
             "type": "enabled",
             "budget_tokens": budget,
@@ -302,6 +361,46 @@ def _ensure_thinking_for_clear_thinking_context_management(
         )
         return True
 
+    @staticmethod
+    def _resolve_clear_thinking_budget_tokens(budget_tokens: int | None) -> int:
+        """Honor an explicit ``budget_tokens`` (including ``0``); only fall back to
+        the Bedrock minimum when the caller omitted it. A truthiness check would
+        wrongly treat an explicit ``0`` as missing."""
+        if budget_tokens is None:
+            return BEDROCK_MIN_THINKING_BUDGET_TOKENS
+        return int(budget_tokens)
+
+    @staticmethod
+    def _effort_from_thinking_budget(budget_tokens: int) -> str:
+        if budget_tokens >= DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET:
+            return "xhigh"
+        if budget_tokens >= DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET:
+            return "high"
+        if budget_tokens >= DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET:
+            return "medium"
+        return "low"
+
+    def _inject_adaptive_thinking_for_clear_thinking(
+        self, anthropic_messages_request: dict, budget_tokens: int, model: str
+    ) -> None:
+        """Adaptive-thinking models (Opus 4.7/4.8, Fable 5) reject
+        ``thinking.type=enabled`` on Bedrock. Use ``thinking.type=adaptive`` plus
+        an ``output_config.effort`` derived from the budget so ``clear_thinking``
+        stays valid without the legacy shape."""
+        output_config = anthropic_messages_request.get("output_config")
+        if not isinstance(output_config, dict):
+            output_config = {}
+        output_config.setdefault(
+            "effort", self._effort_from_thinking_budget(budget_tokens)
+        )
+        anthropic_messages_request["output_config"] = output_config
+        anthropic_messages_request["thinking"] = {"type": "adaptive"}
+        verbose_logger.debug(
+            "Bedrock clear_thinking_20251015: injected adaptive thinking with effort=%s for model=%s",
+            output_config.get("effort"),
+            model,
+        )
+
     def _is_claude_opus_4_5(self, model: str) -> bool:
         """
         Check if the model is Claude Opus 4.5.
@@ -584,6 +683,7 @@ def transform_anthropic_messages_request(
             litellm_params=litellm_params,
             headers=headers,
         )
+        self._normalize_system_role_messages_for_bedrock(anthropic_messages_request)
         #########################################################
         ############## BEDROCK Invoke SPECIFIC TRANSFORMATION ###
         #########################################################
diff --git a/litellm/llms/hosted_vllm/chat/transformation.py b/litellm/llms/hosted_vllm/chat/transformation.py
index 7c42e6a9a002..fc7086548c8b 100644
--- a/litellm/llms/hosted_vllm/chat/transformation.py
+++ b/litellm/llms/hosted_vllm/chat/transformation.py
@@ -20,6 +20,9 @@
     _get_image_mime_type_from_url,
 )
 from litellm.litellm_core_utils.prompt_templates.factory import _parse_mime_type
+from litellm.litellm_core_utils.reasoning_effort_utils import (
+    reasoning_effort_from_thinking_budget,
+)
 from litellm.secret_managers.main import get_secret_str
 from litellm.types.llms.openai import (
     AllMessageValues,
@@ -115,15 +118,11 @@ def map_openai_params(
         if thinking is not None and isinstance(thinking, dict):
             if thinking.get("type") == "enabled":
                 if "reasoning_effort" not in non_default_params:
-                    budget_tokens = thinking.get("budget_tokens", 0)
-                    if budget_tokens >= 10000:
-                        non_default_params["reasoning_effort"] = "high"
-                    elif budget_tokens >= 5000:
-                        non_default_params["reasoning_effort"] = "medium"
-                    elif budget_tokens >= 2000:
-                        non_default_params["reasoning_effort"] = "low"
-                    else:
-                        non_default_params["reasoning_effort"] = "minimal"
+                    non_default_params["reasoning_effort"] = (
+                        reasoning_effort_from_thinking_budget(
+                            thinking.get("budget_tokens", 0)
+                        )
+                    )
 
         return super().map_openai_params(
             non_default_params, optional_params, model, drop_params
diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json
index 6ebac7efc8df..cf7b1b1033c0 100644
--- a/litellm/model_prices_and_context_window_backup.json
+++ b/litellm/model_prices_and_context_window_backup.json
@@ -999,6 +999,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1029,6 +1030,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1059,6 +1061,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1089,6 +1092,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1119,6 +1123,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1149,6 +1154,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1196,6 +1202,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1228,6 +1235,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1260,6 +1268,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1292,6 +1301,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1620,6 +1630,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1652,6 +1663,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1681,6 +1693,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1710,6 +1723,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1739,6 +1753,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1768,6 +1783,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1797,6 +1813,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -2316,6 +2333,7 @@
         "cache_creation_input_token_cost": 6.25e-06,
         "cache_creation_input_token_cost_above_1hr": 1e-05,
         "cache_read_input_token_cost": 5e-07,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -2344,6 +2362,7 @@
         "cache_creation_input_token_cost": 6.25e-06,
         "cache_creation_input_token_cost_above_1hr": 1e-05,
         "cache_read_input_token_cost": 5e-07,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -2470,6 +2489,7 @@
         "max_tokens": 64000,
         "mode": "chat",
         "output_cost_per_token": 1.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -19247,6 +19267,7 @@
         "supported_endpoints": [
             "/v1/chat/completions"
         ],
+        "supports_adaptive_thinking": true,
         "supports_function_calling": true,
         "supports_parallel_function_calling": true,
         "supports_vision": true
@@ -28468,6 +28489,7 @@
         "output_cost_per_token": 1.5e-05,
         "output_cost_per_token_above_200k_tokens": 2.25e-05,
         "source": "https://openrouter.ai/anthropic/claude-sonnet-4.6",
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -28506,6 +28528,7 @@
         "max_tokens": 128000,
         "mode": "chat",
         "output_cost_per_token": 2.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -28566,6 +28589,7 @@
         "max_tokens": 128000,
         "mode": "chat",
         "output_cost_per_token": 2.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -30539,6 +30563,7 @@
     "perplexity/anthropic/claude-opus-4-6": {
         "litellm_provider": "perplexity",
         "mode": "responses",
+        "supports_adaptive_thinking": true,
         "supports_web_search": true,
         "supports_reasoning": false,
         "supports_function_calling": true,
@@ -30547,6 +30572,7 @@
     "perplexity/anthropic/claude-opus-4-7": {
         "litellm_provider": "perplexity",
         "mode": "responses",
+        "supports_adaptive_thinking": true,
         "supports_web_search": true,
         "supports_reasoning": false,
         "supports_function_calling": true,
@@ -33722,6 +33748,7 @@
         "max_tokens": 64000,
         "mode": "chat",
         "output_cost_per_token": 2.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -34955,6 +34982,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -34983,6 +35011,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -35011,6 +35040,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -35040,6 +35070,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -35211,6 +35242,7 @@
         "max_tokens": 64000,
         "mode": "chat",
         "output_cost_per_token": 1.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -42601,6 +42633,7 @@
         "max_tokens": 64000,
         "mode": "chat",
         "output_cost_per_token": 1.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
diff --git a/litellm/types/utils.py b/litellm/types/utils.py
index 24d6e84fba7b..dd70c3c715e1 100644
--- a/litellm/types/utils.py
+++ b/litellm/types/utils.py
@@ -142,6 +142,7 @@ class ProviderSpecificModelInfo(TypedDict, total=False):
     supports_parallel_function_calling: Optional[bool]
     supports_web_search: Optional[bool]
     supports_reasoning: Optional[bool]
+    supports_adaptive_thinking: Optional[bool]
     supports_url_context: Optional[bool]
     supports_none_reasoning_effort: Optional[bool]
     supports_minimal_reasoning_effort: Optional[bool]
diff --git a/litellm/utils.py b/litellm/utils.py
index e0aa85754732..4c2489c6f886 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -6213,6 +6213,9 @@ def _get_model_info_helper(
                 supports_web_search=_model_info.get("supports_web_search", None),
                 supports_url_context=_model_info.get("supports_url_context", None),
                 supports_reasoning=_model_info.get("supports_reasoning", None),
+                supports_adaptive_thinking=_model_info.get(
+                    "supports_adaptive_thinking", None
+                ),
                 supports_none_reasoning_effort=_model_info.get(
                     "supports_none_reasoning_effort", None
                 ),
diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json
index 5f3f22941473..19368072a497 100644
--- a/model_prices_and_context_window.json
+++ b/model_prices_and_context_window.json
@@ -999,6 +999,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1029,6 +1030,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1059,6 +1061,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1089,6 +1092,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1119,6 +1123,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1149,6 +1154,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1196,6 +1202,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1228,6 +1235,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1260,6 +1268,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1292,6 +1301,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1620,6 +1630,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1652,6 +1663,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1681,6 +1693,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1710,6 +1723,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1739,6 +1753,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1768,6 +1783,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -1797,6 +1813,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -2316,6 +2333,7 @@
         "cache_creation_input_token_cost": 6.25e-06,
         "cache_creation_input_token_cost_above_1hr": 1e-05,
         "cache_read_input_token_cost": 5e-07,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -2344,6 +2362,7 @@
         "cache_creation_input_token_cost": 6.25e-06,
         "cache_creation_input_token_cost_above_1hr": 1e-05,
         "cache_read_input_token_cost": 5e-07,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -2470,6 +2489,7 @@
         "max_tokens": 64000,
         "mode": "chat",
         "output_cost_per_token": 1.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -19255,6 +19275,7 @@
         "supported_endpoints": [
             "/v1/chat/completions"
         ],
+        "supports_adaptive_thinking": true,
         "supports_function_calling": true,
         "supports_parallel_function_calling": true,
         "supports_vision": true
@@ -28476,6 +28497,7 @@
         "output_cost_per_token": 1.5e-05,
         "output_cost_per_token_above_200k_tokens": 2.25e-05,
         "source": "https://openrouter.ai/anthropic/claude-sonnet-4.6",
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -28514,6 +28536,7 @@
         "max_tokens": 128000,
         "mode": "chat",
         "output_cost_per_token": 2.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -28574,6 +28597,7 @@
         "max_tokens": 128000,
         "mode": "chat",
         "output_cost_per_token": 2.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -30547,6 +30571,7 @@
     "perplexity/anthropic/claude-opus-4-6": {
         "litellm_provider": "perplexity",
         "mode": "responses",
+        "supports_adaptive_thinking": true,
         "supports_web_search": true,
         "supports_reasoning": false,
         "supports_function_calling": true,
@@ -30555,6 +30580,7 @@
     "perplexity/anthropic/claude-opus-4-7": {
         "litellm_provider": "perplexity",
         "mode": "responses",
+        "supports_adaptive_thinking": true,
         "supports_web_search": true,
         "supports_reasoning": false,
         "supports_function_calling": true,
@@ -33770,6 +33796,7 @@
         "max_tokens": 64000,
         "mode": "chat",
         "output_cost_per_token": 2.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -35003,6 +35030,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -35031,6 +35059,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -35059,6 +35088,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -35088,6 +35118,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": false,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -35259,6 +35290,7 @@
         "max_tokens": 64000,
         "mode": "chat",
         "output_cost_per_token": 1.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
@@ -42811,6 +42843,7 @@
         "max_tokens": 64000,
         "mode": "chat",
         "output_cost_per_token": 1.5e-05,
+        "supports_adaptive_thinking": true,
         "supports_assistant_prefill": true,
         "supports_computer_use": true,
         "supports_function_calling": true,
diff --git a/tests/llm_translation/base_llm_unit_tests.py b/tests/llm_translation/base_llm_unit_tests.py
index a184798b5037..6d845f4b2f14 100644
--- a/tests/llm_translation/base_llm_unit_tests.py
+++ b/tests/llm_translation/base_llm_unit_tests.py
@@ -1458,9 +1458,11 @@ def test_reasoning_effort(self):
             reasoning_effort="high",
         )
         # either accepts reasoning effort or thinking budget
-        assert "reasoning_effort" in optional_params or "4096" in json.dumps(
-            optional_params
-        )
+        from litellm.constants import DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET
+
+        assert "reasoning_effort" in optional_params or str(
+            DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET
+        ) in json.dumps(optional_params)
 
         try:
             litellm._turn_on_debug()
@@ -1777,12 +1779,17 @@ def test_anthropic_reasoning_effort_thinking_translation(self):
             model=base_completion_call_args["model"]
         )
 
+        from litellm.constants import DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET
+
         optional_params = get_optional_params(
             model=base_completion_call_args.get("model"),
             custom_llm_provider=provider,
             reasoning_effort="high",
         )
-        assert optional_params["thinking"] == {"type": "enabled", "budget_tokens": 4096}
+        assert optional_params["thinking"] == {
+            "type": "enabled",
+            "budget_tokens": DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+        }
 
         assert "reasoning_effort" not in optional_params
 
diff --git a/tests/llm_translation/reasoning_effort_grid/grid_spec.py b/tests/llm_translation/reasoning_effort_grid/grid_spec.py
index 9778e01eb975..1bb468d15ad1 100644
--- a/tests/llm_translation/reasoning_effort_grid/grid_spec.py
+++ b/tests/llm_translation/reasoning_effort_grid/grid_spec.py
@@ -1,6 +1,16 @@
 from dataclasses import dataclass, field
 from typing import Dict, FrozenSet, List, Optional, Tuple
 
+from litellm.constants import (
+    ANTHROPIC_MIN_THINKING_BUDGET_TOKENS,
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MAX_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MINIMAL_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET,
+)
+
 OMIT = object()
 
 
@@ -44,12 +54,15 @@ def params(self) -> Dict[str, str]:
 )
 
 _BUDGET_TOKENS: Dict[str, int] = {
-    "minimal": 1024,
-    "low": 1024,
-    "medium": 2048,
-    "high": 4096,
-    "xhigh": 8192,
-    "max": 16384,
+    "minimal": max(
+        DEFAULT_REASONING_EFFORT_MINIMAL_THINKING_BUDGET,
+        ANTHROPIC_MIN_THINKING_BUDGET_TOKENS,
+    ),
+    "low": DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
+    "medium": DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+    "high": DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    "xhigh": DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET,
+    "max": DEFAULT_REASONING_EFFORT_MAX_THINKING_BUDGET,
 }
 
 _ADAPTIVE_EFFORT_LABEL: Dict[str, str] = {
@@ -71,6 +84,10 @@ def params(self) -> Dict[str, str]:
 
 _BAD_REQUEST_EFFORTS: FrozenSet[str] = frozenset({"disabled", "invalid", ""})
 
+# Live providers reject ``max_tokens <= thinking.budget_tokens``, so a budget-mode
+# request must leave room above the largest 200-expected tier (``high``).
+BUDGET_MODE_MAX_TOKENS: int = DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET * 2
+
 
 def _bedrock_clamps_effort(model: "ModelEntry", effort: str) -> bool:
     """Whether Bedrock will clamp ``effort`` down to ``bedrock_effort_ceiling``.
@@ -89,7 +106,9 @@ def _bedrock_clamps_effort(model: "ModelEntry", effort: str) -> bool:
 def expected(model: ModelEntry, effort: str) -> CellExpectation:
     if effort in ("__omit__", "none"):
         if model.mode == "budget":
-            return CellExpectation(status=200, thinking_type=OMIT, max_tokens=8192)
+            return CellExpectation(
+                status=200, thinking_type=OMIT, max_tokens=BUDGET_MODE_MAX_TOKENS
+            )
         return CellExpectation(status=200, thinking_type=OMIT)
 
     if effort in _BAD_REQUEST_EFFORTS:
@@ -117,7 +136,7 @@ def expected(model: ModelEntry, effort: str) -> CellExpectation:
         status=200,
         thinking_type="enabled",
         thinking_budget_tokens=_BUDGET_TOKENS[effort],
-        max_tokens=8192,
+        max_tokens=BUDGET_MODE_MAX_TOKENS,
     )
 
 
diff --git a/tests/llm_translation/reasoning_effort_grid/test_reasoning_effort_grid.py b/tests/llm_translation/reasoning_effort_grid/test_reasoning_effort_grid.py
index a5f16f928e51..304743b1f3cf 100644
--- a/tests/llm_translation/reasoning_effort_grid/test_reasoning_effort_grid.py
+++ b/tests/llm_translation/reasoning_effort_grid/test_reasoning_effort_grid.py
@@ -8,6 +8,7 @@
 from litellm.exceptions import BadRequestError
 
 from .grid_spec import (
+    BUDGET_MODE_MAX_TOKENS,
     OMIT,
     ROUTES,
     CellExpectation,
@@ -28,7 +29,7 @@ def _required_env_missing(model: ModelEntry) -> Optional[str]:
 
 
 def _max_tokens_for(model: ModelEntry) -> int:
-    return 200 if model.mode == "adaptive" else 8192
+    return 200 if model.mode == "adaptive" else BUDGET_MODE_MAX_TOKENS
 
 
 def _build_completion_kwargs(model: ModelEntry, effort: str) -> Dict[str, Any]:
diff --git a/tests/llm_translation/test_optional_params.py b/tests/llm_translation/test_optional_params.py
index 93acf0168337..9ebdb4b7e97c 100644
--- a/tests/llm_translation/test_optional_params.py
+++ b/tests/llm_translation/test_optional_params.py
@@ -1834,12 +1834,17 @@ def test_azure_response_format_param():
     ],
 )
 def test_anthropic_unified_reasoning_content(model, provider):
+    from litellm.constants import DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET
+
     optional_params = get_optional_params(
         model=model,
         custom_llm_provider=provider,
         reasoning_effort="high",
     )
-    assert optional_params["thinking"] == {"type": "enabled", "budget_tokens": 4096}
+    assert optional_params["thinking"] == {
+        "type": "enabled",
+        "budget_tokens": DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    }
 
 
 def test_azure_response_format(monkeypatch):
diff --git a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py
index b111b65e3afb..67eb5b121075 100644
--- a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py
+++ b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py
@@ -9,7 +9,14 @@
 from unittest.mock import MagicMock, patch
 
 import litellm
-from litellm.constants import RESPONSE_FORMAT_TOOL_NAME
+from litellm.constants import (
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MAX_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET,
+    RESPONSE_FORMAT_TOOL_NAME,
+)
 from litellm.llms.anthropic.chat.transformation import AnthropicConfig
 from litellm.llms.anthropic.experimental_pass_through.messages.transformation import (
     AnthropicMessagesConfig,
@@ -2493,9 +2500,9 @@ def test_reasoning_effort_maps_to_budget_thinking_for_non_opus_4_6():
 
     # ``minimal`` floors at ANTHROPIC_MIN_THINKING_BUDGET_TOKENS (1024).
     test_cases = [
-        ("low", 1024),
-        ("medium", 2048),
-        ("high", 4096),
+        ("low", DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET),
+        ("medium", DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET),
+        ("high", DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET),
         ("minimal", 1024),
     ]
 
@@ -2820,7 +2827,10 @@ def test_reasoning_effort_garbage_raises_bad_request(effort):
 
 @pytest.mark.parametrize(
     "effort,expected_budget",
-    [("xhigh", 8192), ("max", 16384)],
+    [
+        ("xhigh", DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET),
+        ("max", DEFAULT_REASONING_EFFORT_MAX_THINKING_BUDGET),
+    ],
 )
 def test_reasoning_effort_xhigh_max_maps_to_budget_on_budget_model(
     effort, expected_budget
diff --git a/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py b/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py
index a58873f5387f..7bcaf07c5bb3 100644
--- a/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py
+++ b/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py
@@ -270,9 +270,9 @@ def test_openai_model_with_thinking_converts_to_reasoning():
             "reasoning" in call_kwargs
         ), "reasoning should be passed to litellm.responses"
 
-        # budget_tokens=1024 -> effort="minimal" (< 2000 threshold)
+        # budget_tokens=1024 -> effort="low" (at the LOW budget threshold)
         # reasoning_auto_summary is False by default, so no summary key
-        expected_reasoning = {"effort": "minimal"}
+        expected_reasoning = {"effort": "low"}
         assert call_kwargs["reasoning"] == expected_reasoning, (
             f"reasoning should be {expected_reasoning} for budget_tokens=1024, "
             f"got {call_kwargs.get('reasoning')}"
@@ -316,7 +316,7 @@ def test_non_claude_model_converts_thinking_to_reasoning_effort(self):
         )
 
         # reasoning_auto_summary is False by default, so no summary key
-        assert result == {"reasoning_effort": "minimal"}
+        assert result == {"reasoning_effort": "low"}
         assert "thinking" not in result
         assert "summary" not in str(result["reasoning_effort"])
 
@@ -336,7 +336,7 @@ def test_translate_thinking_for_model_summary_when_enabled(self):
                 model="openai/gpt-5.2",
             )
             assert result == {
-                "reasoning_effort": {"effort": "medium", "summary": "detailed"}
+                "reasoning_effort": {"effort": "high", "summary": "detailed"}
             }
         finally:
             litellm.reasoning_auto_summary = original
@@ -549,7 +549,7 @@ def test_responses_adapter_preserves_summary(self):
         result = LiteLLMAnthropicToResponsesAPIAdapter.translate_thinking_to_reasoning(
             thinking
         )
-        assert result == {"effort": "medium", "summary": "concise"}
+        assert result == {"effort": "high", "summary": "concise"}
 
     def test_responses_adapter_no_summary_by_default(self):
         """translate_thinking_to_reasoning should not include summary by default (opt-in)."""
@@ -567,7 +567,7 @@ def test_responses_adapter_no_summary_by_default(self):
                     thinking
                 )
             )
-            assert result == {"effort": "medium"}
+            assert result == {"effort": "high"}
             assert result is not None and "summary" not in result
         finally:
             litellm.reasoning_auto_summary = original
@@ -584,7 +584,7 @@ def test_translate_thinking_for_model_preserves_summary(self):
             model="openai/gpt-5.2",
         )
         assert result == {
-            "reasoning_effort": {"effort": "medium", "summary": "concise"}
+            "reasoning_effort": {"effort": "high", "summary": "concise"}
         }
 
 
diff --git a/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_reasoning_effort_translation.py b/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_reasoning_effort_translation.py
index 71755e6da3f4..e3f0bbbcc69c 100644
--- a/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_reasoning_effort_translation.py
+++ b/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_reasoning_effort_translation.py
@@ -3,6 +3,11 @@
 import pytest
 
 import litellm
+from litellm.constants import (
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET,
+)
 from litellm.llms.anthropic.common_utils import AnthropicError
 from litellm.llms.anthropic.experimental_pass_through.messages.transformation import (
     AnthropicMessagesConfig,
@@ -152,7 +157,7 @@ def test_reasoning_effort_unsupported_tier_raises_400_messages(model, bad_effort
     ],
 )
 def test_bedrock_invoke_messages_clamps_effort_to_ceiling(
-    model, effort, expected_effort
+    local_model_cost_map, model, effort, expected_effort
 ):
     """Bedrock Invoke /v1/messages degrades effort to the model's ceiling.
 
@@ -174,7 +179,7 @@ def test_bedrock_invoke_messages_clamps_effort_to_ceiling(
     assert result["thinking"]["type"] == "adaptive"
 
 
-def test_bedrock_invoke_messages_rejects_xhigh_without_ceiling():
+def test_bedrock_invoke_messages_rejects_xhigh_without_ceiling(local_model_cost_map):
     """Sonnet 4.6 on Bedrock has no effort ceiling, so xhigh is still rejected."""
     config = AmazonAnthropicClaudeMessagesConfig()
     optional_params = {"max_tokens": 1024, "reasoning_effort": "xhigh"}
@@ -199,7 +204,9 @@ def test_bedrock_invoke_messages_rejects_xhigh_without_ceiling():
         "bedrock/invoke/us.anthropic.claude-sonnet-4-6",
     ],
 )
-def test_reasoning_effort_max_accepted_on_sonnet_46_messages(model):
+def test_reasoning_effort_max_accepted_on_sonnet_46_messages(
+    local_model_cost_map, model
+):
     config = AnthropicMessagesConfig()
     optional_params = {"max_tokens": 1024, "reasoning_effort": "max"}
 
@@ -269,11 +276,13 @@ def test_reasoning_effort_in_supported_params():
         "bedrock/invoke/us.anthropic.claude-sonnet-4-6",
         "vertex_ai/claude-sonnet-4-6",
         "claude-opus-4-6",
-        "bedrock/invoke/us.anthropic.claude-opus-4-6",
+        "bedrock/invoke/us.anthropic.claude-opus-4-6-v1:0",
         "vertex_ai/claude-opus-4-6",
     ],
 )
-def test_legacy_thinking_high_budget_clamps_to_high_when_xhigh_unsupported(model):
+def test_legacy_thinking_high_budget_clamps_to_high_when_xhigh_unsupported(
+    local_model_cost_map, model
+):
     """Claude Code sends ``thinking.budget_tokens=31999``; Sonnet 4.6 and Opus 4.6
     have no ``xhigh`` tier, so the translator must emit ``high`` rather than the
     provider-invalid ``xhigh`` (regression for issue #29282)."""
@@ -352,13 +361,13 @@ def test_legacy_thinking_translates_to_adaptive_for_opus_48(
 @pytest.mark.parametrize(
     "budget_tokens,expected_effort",
     [
-        (31999, "high"),
-        (24000, "high"),
-        (10000, "high"),
-        (9999, "medium"),
-        (5000, "medium"),
-        (4999, "low"),
-        (1024, "low"),
+        (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET * 2, "high"),
+        (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET, "high"),
+        (DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET, "high"),
+        (DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET - 1, "medium"),
+        (DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET, "medium"),
+        (DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET - 1, "low"),
+        (1, "low"),
     ],
 )
 def test_legacy_thinking_budget_buckets_on_sonnet_46(budget_tokens, expected_effort):
diff --git a/tests/test_litellm/llms/anthropic/experimental_pass_through/responses_adapters/test_responses_adapters_transformation.py b/tests/test_litellm/llms/anthropic/experimental_pass_through/responses_adapters/test_responses_adapters_transformation.py
index 02b817cd334d..14e872485ecf 100644
--- a/tests/test_litellm/llms/anthropic/experimental_pass_through/responses_adapters/test_responses_adapters_transformation.py
+++ b/tests/test_litellm/llms/anthropic/experimental_pass_through/responses_adapters/test_responses_adapters_transformation.py
@@ -11,6 +11,11 @@
 
 sys.path.insert(0, os.path.abspath("../../../../../../.."))
 
+from litellm.constants import (
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+)
 from litellm.llms.anthropic.experimental_pass_through.responses_adapters.transformation import (
     LiteLLMAnthropicToResponsesAPIAdapter,
 )
@@ -628,14 +633,20 @@ def test_budget_above_threshold_high_effort(self):
 
     def test_budget_medium_effort(self):
         result = _ADAPTER.translate_thinking_to_reasoning(
-            {"type": "enabled", "budget_tokens": 7500}
+            {
+                "type": "enabled",
+                "budget_tokens": DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+            }
         )
         assert result == {"effort": "medium"}
         assert result is not None and "summary" not in result
 
     def test_budget_low_effort(self):
         result = _ADAPTER.translate_thinking_to_reasoning(
-            {"type": "enabled", "budget_tokens": 3000}
+            {
+                "type": "enabled",
+                "budget_tokens": DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
+            }
         )
         assert result == {"effort": "low"}
         assert result is not None and "summary" not in result
@@ -648,14 +659,28 @@ def test_budget_minimal_effort(self):
         assert result is not None and "summary" not in result
 
     def test_budget_at_exact_thresholds(self):
+        result_high = _ADAPTER.translate_thinking_to_reasoning(
+            {
+                "type": "enabled",
+                "budget_tokens": DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+            }
+        )
+        assert result_high is not None
+        assert result_high["effort"] == "high"
         result_medium = _ADAPTER.translate_thinking_to_reasoning(
-            {"type": "enabled", "budget_tokens": 5000}
+            {
+                "type": "enabled",
+                "budget_tokens": DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+            }
         )
         assert result_medium is not None
         assert result_medium["effort"] == "medium"
         assert "summary" not in result_medium
         result_low = _ADAPTER.translate_thinking_to_reasoning(
-            {"type": "enabled", "budget_tokens": 2000}
+            {
+                "type": "enabled",
+                "budget_tokens": DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
+            }
         )
         assert result_low is not None
         assert result_low["effort"] == "low"
@@ -670,7 +695,7 @@ def test_non_dict_returns_none(self):
         assert result is None
 
     def test_missing_budget_defaults_to_minimal(self):
-        """Missing budget_tokens defaults to 0, which is < 2000 -> minimal."""
+        """Missing budget_tokens defaults to 0, below the low threshold -> minimal."""
         result = _ADAPTER.translate_thinking_to_reasoning({"type": "enabled"})
         assert result == {"effort": "minimal"}
         assert result is not None and "summary" not in result
@@ -698,7 +723,10 @@ def test_summary_added_when_env_var_set(self):
             litellm.reasoning_auto_summary = False
             os.environ["LITELLM_REASONING_AUTO_SUMMARY"] = "true"
             result = _ADAPTER.translate_thinking_to_reasoning(
-                {"type": "enabled", "budget_tokens": 5000}
+                {
+                    "type": "enabled",
+                    "budget_tokens": DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+                }
             )
             assert result == {"effort": "medium", "summary": "detailed"}
         finally:
diff --git a/tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py b/tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py
index 03b72c504b8f..315e734ef1d4 100644
--- a/tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py
+++ b/tests/test_litellm/llms/anthropic/test_anthropic_common_utils.py
@@ -1465,6 +1465,8 @@ class TestClaudeOpus48AdaptiveThinking:
             "bedrock/eu.anthropic.claude-opus-4-8",
             "vertex_ai/claude-opus-4-8",
             "azure_ai/claude-opus-4-8",
+            "anthropic.claude-opus-4-8-20251201-v1:0",
+            "bedrock/invoke/global.anthropic.claude-opus-4-8-20251201-v1:0",
         ],
     )
     def test_adaptive_thinking_detected_for_opus_4_8(self, local_model_cost_map, model):
@@ -1488,6 +1490,74 @@ def test_resolver_reads_flag_through_bedrock_invoke_prefix(
             is True
         )
 
+    @pytest.mark.parametrize(
+        "model",
+        [
+            "claude-fable-5",
+            "anthropic.claude-fable-5",
+            "us.anthropic.claude-fable-5",
+            "bedrock/invoke/us.anthropic.claude-fable-5",
+            "vertex_ai/claude-fable-5",
+        ],
+    )
+    def test_adaptive_thinking_detected_for_fable_5(self, local_model_cost_map, model):
+        from litellm.llms.anthropic.common_utils import AnthropicModelInfo
+
+        assert AnthropicModelInfo._is_adaptive_thinking_model(model) is True
+
+    @pytest.mark.parametrize(
+        "model",
+        [
+            "claude-opus-4-6",
+            "us.anthropic.claude-opus-4-7",
+            "bedrock/invoke/us.anthropic.claude-opus-4-7",
+            "bedrock/invoke/global.anthropic.claude-opus-4-7-v1:0",
+            "global.anthropic.claude-sonnet-4-6-v1:0",
+            "bedrock/invoke/us.anthropic.claude-opus-4-6-v1:0",
+            "anthropic.claude-opus-4-6-v1",
+            "bedrock/us.anthropic.claude-sonnet-4-6",
+            "us.anthropic.claude-sonnet-4-6",
+            "vertex_ai/claude-opus-4-6",
+            "azure_ai/claude-sonnet-4-6",
+            "claude-sonnet-4-6-20260219",
+            "us.anthropic.claude-sonnet-4-6-20251101-v1:0",
+            "bedrock/invoke/us.anthropic.claude-sonnet-4-6-20251101-v1:0",
+            "claude-sonnet-4.6",
+        ],
+    )
+    def test_adaptive_thinking_detected_for_opus_4_6_4_7_and_sonnet_4_6(
+        self, local_model_cost_map, model
+    ):
+        """Opus 4.6/4.7 and Sonnet 4.6 carry the ``supports_adaptive_thinking`` flag,
+        so detection holds purely from the cost map with no name-based version
+        fallback. Each alias form the Bedrock/anthropic paths see resolves to a flagged
+        base entry through candidate normalization: provider/region prefixes, a
+        Bedrock ``-v1:0`` version suffix (stripped fully for 4.7/4.8 keys or to ``-v1``
+        for the 4.6 key), a dated release suffix (``-20260219``), a combined
+        ``-<date>-v1:0`` suffix (the real Bedrock id shape), and a dotted family
+        version (``4.6`` -> ``4-6``)."""
+        from litellm.llms.anthropic.common_utils import AnthropicModelInfo
+
+        assert AnthropicModelInfo._is_adaptive_thinking_model(model) is True
+
+    @pytest.mark.parametrize(
+        "model",
+        [
+            "claude-opus-4-9",
+            "claude-opus-4-8-some-future-suffix",
+            "us.anthropic.claude-fable-5-preview",
+        ],
+    )
+    def test_unmapped_aliases_defer_to_cost_map(self, local_model_cost_map, model):
+        """Detection is sourced solely from the cost map flag: an alias absent from
+        the map (a future release or a preview suffix) is not adaptive until a
+        ``fallback_generalizations`` rule (PR #29718) covers it."""
+        import litellm
+        from litellm.llms.anthropic.common_utils import AnthropicModelInfo
+
+        assert model not in litellm.model_cost
+        assert AnthropicModelInfo._is_adaptive_thinking_model(model) is False
+
     @pytest.mark.parametrize(
         "model",
         ["claude-opus-4-5", "claude-3-7-sonnet", "claude-3-5-haiku-20241022"],
diff --git a/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py b/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py
index c92a9905229f..c3191f5b5cbd 100644
--- a/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py
+++ b/tests/test_litellm/llms/bedrock/messages/invoke_transformations/test_anthropic_claude3_transformation.py
@@ -18,13 +18,36 @@
     normalize_tool_input_schema_types_for_bedrock_invoke,
     remove_custom_field_from_tools,
 )
-from litellm.constants import BEDROCK_MIN_THINKING_BUDGET_TOKENS
+from litellm.constants import (
+    BEDROCK_MIN_THINKING_BUDGET_TOKENS,
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET,
+)
 from litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_transformation import (
     AmazonAnthropicClaudeMessagesConfig,
     AmazonAnthropicClaudeMessagesStreamDecoder,
 )
 
 
+@pytest.fixture
+def local_model_cost_map(monkeypatch):
+    """Force the bundled backup cost map so adaptive-thinking detection reads this
+    branch's ``supports_adaptive_thinking`` flags, which the network-fetched
+    ``main`` copy lacks until merge."""
+    import litellm
+
+    original = litellm.model_cost
+    monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True")
+    litellm.model_cost = litellm.get_model_cost_map(url="")
+    litellm.get_model_info.cache_clear()
+    try:
+        yield
+    finally:
+        litellm.model_cost = original
+        litellm.get_model_info.cache_clear()
+
+
 @pytest.mark.asyncio
 async def test_bedrock_sse_wrapper_encodes_dict_chunks():
     """Verify that `bedrock_sse_wrapper` converts dictionary chunks to properly formatted Server-Sent Events and forwards non-dict chunks unchanged."""
@@ -388,7 +411,9 @@ def test_bedrock_invoke_messages_transform_adds_name_when_tool_missing_name():
     assert result["tools"][0]["name"] == "litellm_unnamed_tool_0"
 
 
-def test_bedrock_invoke_messages_skips_thinking_injection_when_already_enabled():
+def test_bedrock_invoke_messages_skips_thinking_injection_when_already_enabled(
+    local_model_cost_map,
+):
     from litellm.types.router import GenericLiteLLMParams
 
     cfg = AmazonAnthropicClaudeMessagesConfig()
@@ -624,9 +649,9 @@ def test_bedrock_messages_strips_output_config():
             headers={},
         )
 
-    assert (
-        "output_config" not in result
-    ), "output_config should be stripped for models that don't support it"
+    assert "output_config" not in result, (
+        "output_config should be stripped for models that don't support it"
+    )
     assert result.get("max_tokens") == 4096
 
 
@@ -659,9 +684,9 @@ def test_bedrock_messages_preserves_output_config_for_claude_4_6():
             headers={},
         )
 
-    assert (
-        "output_config" in result
-    ), "output_config should be preserved for supported models"
+    assert "output_config" in result, (
+        "output_config should be preserved for supported models"
+    )
     assert result["output_config"] == {"effort": "high"}
     assert result.get("max_tokens") == 4096
 
@@ -1038,7 +1063,7 @@ def test_bedrock_messages_drop_params_keeps_output_config_for_4_7():
     ],
 )
 def test_bedrock_messages_maps_reasoning_effort_for_adaptive_model(
-    reasoning_effort, expected_effort
+    local_model_cost_map, reasoning_effort, expected_effort
 ):
     """``reasoning_effort`` maps to ``thinking`` + ``output_config.effort`` on /v1/messages."""
     from unittest.mock import patch
@@ -1204,9 +1229,9 @@ def test_bedrock_messages_strips_context_management():
         headers={},
     )
 
-    assert (
-        "context_management" not in result
-    ), "context_management should be stripped — Bedrock Invoke rejects it"
+    assert "context_management" not in result, (
+        "context_management should be stripped — Bedrock Invoke rejects it"
+    )
     assert result.get("max_tokens") == 4096
 
 
@@ -1353,12 +1378,12 @@ def test_bedrock_messages_filters_user_provided_unsupported_beta_header():
     )
 
     betas = result.get("anthropic_beta") or []
-    assert (
-        "advisor-tool-2026-03-01" not in betas
-    ), "user-provided beta not in the Bedrock mapping must be dropped"
-    assert (
-        "context-1m-2025-08-07" in betas
-    ), "user-provided beta that IS in the Bedrock mapping should survive"
+    assert "advisor-tool-2026-03-01" not in betas, (
+        "user-provided beta not in the Bedrock mapping must be dropped"
+    )
+    assert "context-1m-2025-08-07" in betas, (
+        "user-provided beta that IS in the Bedrock mapping should survive"
+    )
 
 
 def test_bedrock_messages_renames_user_provided_aliased_beta_header():
@@ -1383,12 +1408,12 @@ def test_bedrock_messages_renames_user_provided_aliased_beta_header():
     )
 
     betas = result.get("anthropic_beta") or []
-    assert (
-        "advanced-tool-use-2025-11-20" not in betas
-    ), "Anthropic-direct spelling should be rewritten, not forwarded verbatim"
-    assert (
-        "tool-search-tool-2025-10-19" in betas
-    ), "user-provided beta should be renamed to the Bedrock-side spelling"
+    assert "advanced-tool-use-2025-11-20" not in betas, (
+        "Anthropic-direct spelling should be rewritten, not forwarded verbatim"
+    )
+    assert "tool-search-tool-2025-10-19" in betas, (
+        "user-provided beta should be renamed to the Bedrock-side spelling"
+    )
 
 
 @pytest.mark.asyncio
@@ -1648,3 +1673,339 @@ async def _stream():  # type: ignore[return-type]
         custom_llm_provider="bedrock",
     )
     assert cost == pytest.approx(0.052150725, rel=0, abs=1e-9)
+
+
+@pytest.mark.parametrize(
+    "model",
+    [
+        "us.anthropic.claude-opus-4-8",
+        "global.anthropic.claude-opus-4-7-v1:0",
+        "us.anthropic.claude-fable-5",
+        "global.anthropic.claude-fable-5",
+    ],
+)
+def test_bedrock_clear_thinking_injects_adaptive_with_effort_for_adaptive_models(
+    local_model_cost_map, model
+):
+    """clear_thinking_20251015 without a top-level ``thinking`` field must inject
+    ``thinking.type=adaptive`` plus ``output_config.effort`` on adaptive-thinking
+    models (Opus 4.7/4.8, Fable 5). The legacy ``thinking.type=enabled`` shape is
+    rejected by Bedrock for these models (issue #29188). Detection is sourced from
+    the cost-map ``supports_adaptive_thinking`` flag; the versioned Bedrock id
+    (``-v1:0``) resolves to its suffix-less cost-map entry."""
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    request = {
+        "max_tokens": 32000,
+        "context_management": {
+            "edits": [{"type": "clear_thinking_20251015", "keep": "all"}]
+        },
+    }
+
+    changed = cfg._ensure_thinking_for_clear_thinking_context_management(
+        anthropic_messages_request=request, model=model
+    )
+
+    assert changed is True
+    assert request["thinking"] == {"type": "adaptive"}
+    assert request["output_config"]["effort"] == "low"
+
+
+def test_bedrock_clear_thinking_converts_legacy_enabled_budget_to_effort():
+    """A legacy ``thinking.type=enabled`` with a budget on an adaptive model is
+    translated to ``adaptive`` + ``output_config.effort`` derived from the budget
+    rather than forwarded as the rejected ``enabled`` shape."""
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    request = {
+        "max_tokens": 32000,
+        "thinking": {
+            "type": "enabled",
+            "budget_tokens": DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+        },
+        "context_management": {
+            "edits": [{"type": "clear_thinking_20251015", "keep": "all"}]
+        },
+    }
+
+    changed = cfg._ensure_thinking_for_clear_thinking_context_management(
+        anthropic_messages_request=request, model="us.anthropic.claude-opus-4-8"
+    )
+
+    assert changed is True
+    assert request["thinking"] == {"type": "adaptive"}
+    assert request["output_config"]["effort"] == "high"
+
+
+def test_resolve_clear_thinking_budget_tokens_honors_explicit_zero():
+    """A truthiness guard would treat an explicit ``budget_tokens=0`` as missing
+    and silently substitute the Bedrock minimum. The resolver must honor ``0``
+    and only fall back to the minimum when the caller omits the budget."""
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    assert cfg._resolve_clear_thinking_budget_tokens(0) == 0
+    assert (
+        cfg._resolve_clear_thinking_budget_tokens(None)
+        == BEDROCK_MIN_THINKING_BUDGET_TOKENS
+    )
+    assert cfg._resolve_clear_thinking_budget_tokens(12000) == 12000
+
+
+def test_bedrock_clear_thinking_keeps_enabled_for_non_adaptive_models():
+    """Non-adaptive extended-thinking models (e.g. Opus 4.5) still use the legacy
+    ``thinking.type=enabled`` + ``budget_tokens`` shape, which they accept."""
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    request = {
+        "max_tokens": 32000,
+        "context_management": {
+            "edits": [{"type": "clear_thinking_20251015", "keep": "all"}]
+        },
+    }
+
+    changed = cfg._ensure_thinking_for_clear_thinking_context_management(
+        anthropic_messages_request=request, model="anthropic.claude-opus-4-5-v1:0"
+    )
+
+    assert changed is True
+    assert request["thinking"] == {
+        "type": "enabled",
+        "budget_tokens": BEDROCK_MIN_THINKING_BUDGET_TOKENS,
+    }
+    assert "output_config" not in request
+
+
+def test_bedrock_invoke_transform_emits_adaptive_thinking_for_opus_4_8():
+    """End-to-end: a Claude Code clear_thinking payload (no ``thinking`` field) on
+    Opus 4.8 must produce a Bedrock Invoke body with adaptive thinking and an
+    effort, never ``thinking.type.enabled``."""
+    from litellm.types.router import GenericLiteLLMParams
+
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    optional_params = {
+        "max_tokens": 32000,
+        "stream": False,
+        "context_management": {
+            "edits": [{"type": "clear_thinking_20251015", "keep": "all"}]
+        },
+    }
+
+    result = cfg.transform_anthropic_messages_request(
+        model="us.anthropic.claude-opus-4-8",
+        messages=[{"role": "user", "content": "hi"}],
+        anthropic_messages_optional_request_params=copy.deepcopy(optional_params),
+        litellm_params=GenericLiteLLMParams(),
+        headers={},
+    )
+
+    assert result["thinking"] == {"type": "adaptive"}
+    assert result["output_config"]["effort"] == "low"
+
+
+def test_bedrock_invoke_transform_normalizes_system_role_message_into_system():
+    """Bedrock Invoke rejects ``role: "system"`` entries in ``messages`` on some
+    Claude aliases; they must be moved into the top-level ``system`` field
+    (Anthropic Messages carries that content there)."""
+    from litellm.types.router import GenericLiteLLMParams
+
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    messages = [
+        {"role": "system", "content": "You are a careful assistant."},
+        {"role": "user", "content": "hi"},
+    ]
+
+    result = cfg.transform_anthropic_messages_request(
+        model="anthropic.claude-3-haiku-20240307-v1:0",
+        messages=copy.deepcopy(messages),
+        anthropic_messages_optional_request_params={"max_tokens": 256, "stream": False},
+        litellm_params=GenericLiteLLMParams(),
+        headers={},
+    )
+
+    assert all(m.get("role") != "system" for m in result["messages"])
+    assert result["messages"] == [{"role": "user", "content": "hi"}]
+    assert result["system"] == [
+        {"type": "text", "text": "You are a careful assistant."}
+    ]
+
+
+def test_bedrock_invoke_transform_merges_system_role_into_existing_system():
+    """A ``role: "system"`` message is appended to any pre-existing top-level
+    ``system`` content rather than replacing it."""
+    from litellm.types.router import GenericLiteLLMParams
+
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    messages = [
+        {"role": "system", "content": "Follow the user's formatting."},
+        {"role": "user", "content": "hi"},
+    ]
+
+    result = cfg.transform_anthropic_messages_request(
+        model="anthropic.claude-3-haiku-20240307-v1:0",
+        messages=copy.deepcopy(messages),
+        anthropic_messages_optional_request_params={
+            "max_tokens": 256,
+            "stream": False,
+            "system": "Base system prompt.",
+        },
+        litellm_params=GenericLiteLLMParams(),
+        headers={},
+    )
+
+    assert result["messages"] == [{"role": "user", "content": "hi"}]
+    assert result["system"] == [
+        {"type": "text", "text": "Base system prompt."},
+        {"type": "text", "text": "Follow the user's formatting."},
+    ]
+
+
+def test_bedrock_invoke_transform_merges_list_content_system_role_into_system():
+    """A ``role: "system"`` message whose content is a list of content blocks is
+    merged block-by-block into the top-level ``system`` field, preserving any
+    pre-existing system blocks."""
+    from litellm.types.router import GenericLiteLLMParams
+
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    messages = [
+        {
+            "role": "system",
+            "content": [
+                {"type": "text", "text": "Be terse."},
+                {"type": "text", "text": "Cite sources."},
+            ],
+        },
+        {"role": "user", "content": "hi"},
+    ]
+
+    result = cfg.transform_anthropic_messages_request(
+        model="anthropic.claude-3-haiku-20240307-v1:0",
+        messages=copy.deepcopy(messages),
+        anthropic_messages_optional_request_params={
+            "max_tokens": 256,
+            "stream": False,
+            "system": [{"type": "text", "text": "Base."}],
+        },
+        litellm_params=GenericLiteLLMParams(),
+        headers={},
+    )
+
+    assert result["messages"] == [{"role": "user", "content": "hi"}]
+    assert result["system"] == [
+        {"type": "text", "text": "Base."},
+        {"type": "text", "text": "Be terse."},
+        {"type": "text", "text": "Cite sources."},
+    ]
+
+
+def test_as_system_content_blocks_handles_each_shape():
+    """``_as_system_content_blocks`` normalizes every system shape: ``None`` -> empty,
+    a string -> a single text block, a list -> a shallow copy, and any other value
+    (e.g. a bare content-block dict) -> wrapped in a single-element list."""
+    block = {"type": "text", "text": "x"}
+    assert AmazonAnthropicClaudeMessagesConfig._as_system_content_blocks(None) == []
+    assert AmazonAnthropicClaudeMessagesConfig._as_system_content_blocks("hello") == [
+        {"type": "text", "text": "hello"}
+    ]
+    blocks = [block]
+    out = AmazonAnthropicClaudeMessagesConfig._as_system_content_blocks(blocks)
+    assert out == blocks and out is not blocks
+    assert AmazonAnthropicClaudeMessagesConfig._as_system_content_blocks(block) == [
+        block
+    ]
+
+
+@pytest.mark.parametrize(
+    "budget_tokens,expected_effort",
+    [
+        (0, "low"),
+        (DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET - 1, "low"),
+        (DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET, "medium"),
+        (DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET - 1, "medium"),
+        (DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET, "high"),
+        (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET - 1, "high"),
+        (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET, "xhigh"),
+        (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET * 2, "xhigh"),
+    ],
+)
+def test_effort_from_thinking_budget_tiers(budget_tokens, expected_effort):
+    """The budget -> effort mapping pins each tier boundary so a shifted threshold
+    is caught."""
+    assert (
+        AmazonAnthropicClaudeMessagesConfig._effort_from_thinking_budget(budget_tokens)
+        == expected_effort
+    )
+
+
+def test_inject_adaptive_thinking_preserves_existing_effort():
+    """When the request already carries an ``output_config.effort`` we keep it
+    rather than overwriting with the budget-derived tier."""
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    request = {"output_config": {"effort": "max", "other": "keep"}}
+
+    cfg._inject_adaptive_thinking_for_clear_thinking(
+        request, budget_tokens=24000, model="us.anthropic.claude-fable-5"
+    )
+
+    assert request["thinking"] == {"type": "adaptive"}
+    assert request["output_config"] == {"effort": "max", "other": "keep"}
+
+
+def test_bedrock_clear_thinking_noops_when_thinking_already_adaptive():
+    """An incoming ``thinking.type=adaptive`` is already valid for clear_thinking;
+    the helper must leave the request untouched and report no change."""
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    request = {
+        "max_tokens": 32000,
+        "thinking": {"type": "adaptive"},
+        "context_management": {
+            "edits": [{"type": "clear_thinking_20251015", "keep": "all"}]
+        },
+    }
+
+    changed = cfg._ensure_thinking_for_clear_thinking_context_management(
+        anthropic_messages_request=request, model="us.anthropic.claude-fable-5"
+    )
+
+    assert changed is False
+    assert request["thinking"] == {"type": "adaptive"}
+    assert "output_config" not in request
+
+
+def test_bedrock_clear_thinking_replaces_disabled_thinking_on_adaptive_model():
+    """A ``thinking.type=disabled`` (or any non-enabled/adaptive shape) is invalid
+    for clear_thinking on Bedrock; it must be replaced with a usable config. On an
+    adaptive model that means ``thinking.type=adaptive`` + ``output_config.effort``."""
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    request = {
+        "max_tokens": 32000,
+        "thinking": {"type": "disabled"},
+        "context_management": {
+            "edits": [{"type": "clear_thinking_20251015", "keep": "all"}]
+        },
+    }
+
+    changed = cfg._ensure_thinking_for_clear_thinking_context_management(
+        anthropic_messages_request=request, model="us.anthropic.claude-fable-5"
+    )
+
+    assert changed is True
+    assert request["thinking"] == {"type": "adaptive"}
+    assert request["output_config"]["effort"] == "low"
+
+
+def test_bedrock_clear_thinking_leaves_enabled_thinking_on_non_adaptive_model():
+    """A caller-supplied ``thinking.type=enabled`` on a non-adaptive extended-thinking
+    model (Opus 4.5) is already accepted by Bedrock, so the helper leaves it as-is
+    and does not convert it to the adaptive shape."""
+    cfg = AmazonAnthropicClaudeMessagesConfig()
+    request = {
+        "max_tokens": 32000,
+        "thinking": {"type": "enabled", "budget_tokens": 8000},
+        "context_management": {
+            "edits": [{"type": "clear_thinking_20251015", "keep": "all"}]
+        },
+    }
+
+    changed = cfg._ensure_thinking_for_clear_thinking_context_management(
+        anthropic_messages_request=request, model="anthropic.claude-opus-4-5-v1:0"
+    )
+
+    assert changed is False
+    assert request["thinking"] == {"type": "enabled", "budget_tokens": 8000}
+    assert "output_config" not in request
diff --git a/tests/test_litellm/llms/hosted_vllm/chat/test_hosted_vllm_chat_transformation.py b/tests/test_litellm/llms/hosted_vllm/chat/test_hosted_vllm_chat_transformation.py
index 5c1f1dcb63d3..3ddb67b9f8dd 100644
--- a/tests/test_litellm/llms/hosted_vllm/chat/test_hosted_vllm_chat_transformation.py
+++ b/tests/test_litellm/llms/hosted_vllm/chat/test_hosted_vllm_chat_transformation.py
@@ -7,6 +7,10 @@
     0, os.path.abspath("../../../../..")
 )  # Adds the parent directory to the system path
 
+from litellm.constants import (
+    DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+    DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
+)
 from litellm.llms.hosted_vllm.chat.transformation import HostedVLLMChatConfig
 
 
@@ -135,9 +139,14 @@ def test_hosted_vllm_supports_thinking():
     )
     assert "thinking" in supported_params
 
-    # Test thinking with low budget_tokens -> "minimal" (for < 2000)
+    # Test thinking below the low threshold -> "minimal"
     optional_params = config.map_openai_params(
-        non_default_params={"thinking": {"type": "enabled", "budget_tokens": 1024}},
+        non_default_params={
+            "thinking": {
+                "type": "enabled",
+                "budget_tokens": DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET - 1,
+            }
+        },
         optional_params={},
         model="hosted_vllm/GLM-4.6-FP8",
         drop_params=False,
@@ -147,7 +156,12 @@ def test_hosted_vllm_supports_thinking():
 
     # Test thinking with high budget_tokens -> "high"
     optional_params = config.map_openai_params(
-        non_default_params={"thinking": {"type": "enabled", "budget_tokens": 15000}},
+        non_default_params={
+            "thinking": {
+                "type": "enabled",
+                "budget_tokens": DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+            }
+        },
         optional_params={},
         model="hosted_vllm/GLM-4.6-FP8",
         drop_params=False,
@@ -157,7 +171,10 @@ def test_hosted_vllm_supports_thinking():
     # Test that existing reasoning_effort is not overwritten
     optional_params = config.map_openai_params(
         non_default_params={
-            "thinking": {"type": "enabled", "budget_tokens": 15000},
+            "thinking": {
+                "type": "enabled",
+                "budget_tokens": DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
+            },
             "reasoning_effort": "low",
         },
         optional_params={},