diff --git a/hermes_cli/runtime_provider.py b/hermes_cli/runtime_provider.py
index 1fe5acc2b659..d625b75ba19c 100644
--- a/hermes_cli/runtime_provider.py
+++ b/hermes_cli/runtime_provider.py
@@ -260,11 +260,16 @@ def _resolve_runtime_from_pool_entry(
             if cfg_base_url:
                 base_url = cfg_base_url
         configured_mode = _parse_api_mode(model_cfg.get("api_mode"))
-        if configured_mode and _provider_supports_explicit_api_mode(provider, configured_provider):
-            api_mode = configured_mode
-        elif provider in ("opencode-zen", "opencode-go"):
+        if provider in ("opencode-zen", "opencode-go"):
+            # Re-derive api_mode from the effective model rather than the
+            # persisted api_mode: the opencode providers serve both
+            # anthropic_messages and chat_completions models, so the previous
+            # session's mode must not leak across /model switches.
+            # Refs #16878.
             from hermes_cli.models import opencode_model_api_mode
             api_mode = opencode_model_api_mode(provider, effective_model)
+        elif configured_mode and _provider_supports_explicit_api_mode(provider, configured_provider):
+            api_mode = configured_mode
         else:
             # Auto-detect Anthropic-compatible endpoints (/anthropic suffix,
             # Kimi /coding, api.openai.com → codex_responses, api.x.ai →
@@ -1212,15 +1217,20 @@ def resolve_runtime_provider(
             configured_provider = str(model_cfg.get("provider") or "").strip().lower()
             # Only honor persisted api_mode when it belongs to the same provider family.
             configured_mode = _parse_api_mode(model_cfg.get("api_mode"))
-            if configured_mode and _provider_supports_explicit_api_mode(provider, configured_provider):
-                api_mode = configured_mode
-            elif provider in ("opencode-zen", "opencode-go"):
+            if provider in ("opencode-zen", "opencode-go"):
+                # opencode-zen/go must always re-derive api_mode from the
+                # target model (not the stale persisted api_mode), because
+                # the same provider serves both anthropic_messages
+                # (e.g. minimax-m2.7) and chat_completions (e.g.
+                # deepseek-v4-flash) and switching models via /model would
+                # otherwise carry the previous mode forward, stripping /v1
+                # from base_url for chat_completions models and 404'ing.
+                # Refs #16878.
                 from hermes_cli.models import opencode_model_api_mode
-                # Prefer the target_model from the caller (explicit mid-session
-                # switch) over the stale model.default; see _resolve_runtime_from_pool_entry
-                # for the same rationale.
                 _effective = target_model or model_cfg.get("default", "")
                 api_mode = opencode_model_api_mode(provider, _effective)
+            elif configured_mode and _provider_supports_explicit_api_mode(provider, configured_provider):
+                api_mode = configured_mode
             else:
                 # Auto-detect Anthropic-compatible endpoints by URL convention
                 # (e.g. https://api.minimax.io/anthropic, https://dashscope.../anthropic)
diff --git a/run_agent.py b/run_agent.py
index 3f2b7830824f..20153e62e47c 100644
--- a/run_agent.py
+++ b/run_agent.py
@@ -8089,16 +8089,23 @@ def _build_assistant_message(self, assistant_message, finish_reason: str) -> dic
             "finish_reason": finish_reason,
         }
 
-        if hasattr(assistant_message, "reasoning_content"):
-            raw_reasoning_content = getattr(assistant_message, "reasoning_content", None)
-            if raw_reasoning_content is not None:
-                msg["reasoning_content"] = _sanitize_surrogates(raw_reasoning_content)
-            elif msg.get("tool_calls") and self._needs_deepseek_tool_reasoning():
-                # DeepSeek thinking mode requires reasoning_content on every
-                # assistant tool-call message. Without it, replaying the
-                # persisted message causes HTTP 400. Include empty string
-                # as a defensive compatibility fallback (refs #15250).
-                msg["reasoning_content"] = ""
+        # Always persist `reasoning_content` on assistant turns. Without this
+        # the persisted history is silently incompatible with DeepSeek/Kimi
+        # thinking-mode replay (HTTP 400: "The reasoning_content in the
+        # thinking mode must be passed back to the API.") whenever the user
+        # later switches model. Prefer the SDK-supplied value (may carry
+        # structured data); otherwise fall back to the already-sanitized
+        # `reasoning_text` we accumulated from streaming deltas; finally
+        # default to "" so non-thinking providers ignore it harmlessly while
+        # thinking providers see a valid (empty) field. Refs #16844 (write
+        # side) / #15213, #15250, #15353, #15741, #15748 (read-side patches).
+        sdk_reasoning_content = getattr(assistant_message, "reasoning_content", None)
+        if sdk_reasoning_content is not None:
+            msg["reasoning_content"] = _sanitize_surrogates(sdk_reasoning_content)
+        elif reasoning_text:
+            msg["reasoning_content"] = reasoning_text
+        else:
+            msg["reasoning_content"] = ""
 
         if hasattr(assistant_message, 'reasoning_details') and assistant_message.reasoning_details:
             # Pass reasoning_details back unmodified so providers (OpenRouter,