test(e2e): bump claude code compat matrix tiers to sonnet-5 and opus-4-8

tests/e2e/claude_code/_builder_unit_tests/fixtures/expected_matrix.json

@@ -26,7 +26,7 @@
       "providers": {
         "anthropic": {
           "status": "fail",
-          "error": "[claude-sonnet-4-6] tool call dropped"
+          "error": "[claude-sonnet-5] tool call dropped"
         },
         "bedrock_invoke": {
           "status": "not_applicable",

tests/e2e/claude_code/_builder_unit_tests/fixtures/results.json

@@ -10,13 +10,13 @@
     {
       "feature_id": "basic_messaging_non_streaming",
       "provider": "anthropic",
-      "nodeid": "tests/e2e/claude_code/basic_messaging_non_streaming/test_anthropic.py::test_basic_messaging_non_streaming_anthropic[claude-sonnet-4-6]",
+      "nodeid": "tests/e2e/claude_code/basic_messaging_non_streaming/test_anthropic.py::test_basic_messaging_non_streaming_anthropic[claude-sonnet-5]",
       "result": {"status": "pass"}
     },
     {
       "feature_id": "basic_messaging_non_streaming",
       "provider": "anthropic",
-      "nodeid": "tests/e2e/claude_code/basic_messaging_non_streaming/test_anthropic.py::test_basic_messaging_non_streaming_anthropic[claude-opus-4-7]",
+      "nodeid": "tests/e2e/claude_code/basic_messaging_non_streaming/test_anthropic.py::test_basic_messaging_non_streaming_anthropic[claude-opus-4-8]",
       "result": {"status": "pass"}
     },
     {
@@ -28,8 +28,8 @@
     {
       "feature_id": "tool_use",
       "provider": "anthropic",
-      "nodeid": "tests/e2e/claude_code/tool_use/test_anthropic.py::test_x[claude-sonnet-4-6]",
-      "result": {"status": "fail", "error": "[claude-sonnet-4-6] tool call dropped"}
+      "nodeid": "tests/e2e/claude_code/tool_use/test_anthropic.py::test_x[claude-sonnet-5]",
+      "result": {"status": "fail", "error": "[claude-sonnet-5] tool call dropped"}
     },
     {
       "feature_id": "tool_use",

tests/e2e/claude_code/_builder_unit_tests/test_matrix_builder.py

@@ -74,7 +74,7 @@ def test_build_matrix_any_fail_makes_cell_fail():
         {
             "feature_id": "f",
             "provider": "anthropic",
-            "result": {"status": "fail", "error": "[claude-opus-4-7] timeout"},
+            "result": {"status": "fail", "error": "[claude-opus-4-8] timeout"},
         },
         {"feature_id": "f", "provider": "anthropic", "result": {"status": "pass"}},
     ]
@@ -87,7 +87,7 @@ def test_build_matrix_any_fail_makes_cell_fail():
     )
     cell = matrix["features"][0]["providers"]["anthropic"]
     assert cell["status"] == "fail"
-    assert cell["error"] == "[claude-opus-4-7] timeout"
+    assert cell["error"] == "[claude-opus-4-8] timeout"
 
 
 def test_build_matrix_joins_all_failure_errors_in_one_cell():
@@ -110,7 +110,7 @@ def test_build_matrix_joins_all_failure_errors_in_one_cell():
         {
             "feature_id": "f",
             "provider": "anthropic",
-            "result": {"status": "fail", "error": "[claude-opus-4-7] timeout"},
+            "result": {"status": "fail", "error": "[claude-opus-4-8] timeout"},
         },
     ]
     matrix = build_matrix(
@@ -123,7 +123,7 @@ def test_build_matrix_joins_all_failure_errors_in_one_cell():
     cell = matrix["features"][0]["providers"]["anthropic"]
     assert cell["status"] == "fail"
     assert "[claude-haiku-4-5] 429" in cell["error"]
-    assert "[claude-opus-4-7] timeout" in cell["error"]
+    assert "[claude-opus-4-8] timeout" in cell["error"]
 
 
 def test_build_matrix_mixed_pass_and_not_tested_surfaces_pass():
@@ -393,7 +393,7 @@ def test_build_matrix_6x5_grid_matches_published_sample():
 
     feature_ids = [feature["id"] for feature in manifest["features"]]
     providers = manifest["providers"]
-    models = ["claude-haiku-4-5", "claude-sonnet-4-6", "claude-opus-4-7"]
+    models = ["claude-haiku-4-5", "claude-sonnet-5", "claude-opus-4-8"]
 
     results = []
     for feature_id in feature_ids:
@@ -440,7 +440,7 @@ def test_build_matrix_1x5_grid_one_failing_model_breaks_cell():
             "provider": "bedrock_invoke",
             "result": {
                 "status": "fail",
-                "error": "[claude-opus-4-7-bedrock-invoke] claude CLI exited 1: throttled",
+                "error": "[claude-opus-4-8-bedrock-invoke] claude CLI exited 1: throttled",
             },
         },
         {
@@ -459,7 +459,7 @@ def test_build_matrix_1x5_grid_one_failing_model_breaks_cell():
     )
     cell = matrix["features"][0]["providers"]["bedrock_invoke"]
     assert cell["status"] == "fail"
-    assert "claude-opus-4-7-bedrock-invoke" in cell["error"]
+    assert "claude-opus-4-8-bedrock-invoke" in cell["error"]
 
 
 def test_build_from_paths_writes_output(tmp_path):

tests/e2e/claude_code/_builder_unit_tests/test_v0_layout.py

@@ -141,18 +141,37 @@ def test_every_manifest_feature_has_per_provider_test_file(feature_id, provider)
     assert test_file.is_file(), f"missing per-provider test file: {test_file}"
 
 
+DEFAULT_TIERS = ("haiku-4-5", "sonnet-5", "opus-4-8")
+
+PROVIDER_TIER_OVERRIDES = {
+    "vertex_ai": ("haiku-4-5", "sonnet-4-6", "opus-4-7"),
+    "azure": ("haiku-4-5", "sonnet-4-6", "opus-4-8"),
+}
+
+
 @pytest.mark.parametrize("feature_id", EXPECTED_FEATURE_IDS)
 @pytest.mark.parametrize("provider", EXPECTED_PROVIDERS)
 def test_per_provider_test_file_imports_and_parametrizes_three_models(
     feature_id, provider
 ):
     """Every test file must reference the three Claude tiers required
-    by the PRD: Haiku 4.5, Sonnet 4.6, Opus 4.7. Implementations may
-    use plain aliases or per-provider-suffixed aliases (e.g.
-    `claude-opus-4-7-bedrock-invoke`), so we check for the tier
-    substrings rather than exact alias names."""
+    by the PRD: small, mid, and large; Haiku 4.5, Sonnet 5, and Opus
+    4.8 as of the 2026-07 tier bump. Implementations may use plain
+    aliases or per-provider-suffixed aliases (e.g.
+    `claude-opus-4-8-bedrock-invoke`), so we check for the tier
+    substrings rather than exact alias names.
+
+    Two provider columns are pinned to older tiers until quota is
+    granted: Vertex AI has 0 TPM for the sonnet-5 / opus-4-8 base
+    models in the suite's project+region (deterministic 429), and
+    Microsoft Foundry has a hard 0 TPM quota limit for Claude Sonnet 5
+    in the suite's subscription, so its sonnet slot stays on
+    sonnet-4-6 while its opus slot runs the deployed opus-4-8. Once
+    quota requests land, flip the overrides here and the matching
+    aliases in `test_config.yaml` and the per-provider test files."""
+    tiers = PROVIDER_TIER_OVERRIDES.get(provider, DEFAULT_TIERS)
     text = (REPO_ROOT / feature_id / f"test_{provider}.py").read_text()
-    for tier in ("haiku-4-5", "sonnet-4-6", "opus-4-7"):
+    for tier in tiers:
         assert (
             tier in text
         ), f"{feature_id}/test_{provider}.py does not reference {tier}"
@@ -179,5 +198,5 @@ def test_azure_test_file_drives_the_proxy(feature_id):
     )
     assert '"status": "not_applicable"' not in text, (
         f"{feature_id}/test_azure.py still reports not_applicable; Microsoft Foundry "
-        "now hosts Claude (Haiku 4.5, Sonnet 4.6, Opus 4.7), so this row must run."
+        "now hosts Claude (Haiku 4.5, Sonnet 4.6, Opus 4.8), so this row must run."
     )

tests/e2e/claude_code/_driver_unit_tests/test_basic_messaging.py

@@ -184,8 +184,8 @@ def test_verify_streaming_requires_all_models_to_stream(monkeypatch):
     fake_result = _FakeResult()
     outcomes = {
         "claude-haiku-4-5": DriverResult(text="ok", events=_streamed_events(5)),
-        "claude-sonnet-4-6": DriverResult(text="ok", events=_buffered_events()),
-        "claude-opus-4-7": DriverResult(text="ok", events=_streamed_events(5)),
+        "claude-sonnet-5": DriverResult(text="ok", events=_buffered_events()),
+        "claude-opus-4-8": DriverResult(text="ok", events=_streamed_events(5)),
     }
     _install_fake_runner(monkeypatch, outcomes_by_model=outcomes)
 

tests/e2e/claude_code/_driver_unit_tests/test_cli_driver.py

@@ -97,7 +97,7 @@ def test_run_claude_overlays_proxy_env():
     runner, captured = _make_runner(stdout="")
     run_claude(
         prompt="hi",
-        model="claude-opus-4-7",
+        model="claude-opus-4-8",
         base_url="http://proxy.example:4000",
         api_key="sk-abc",
         runner=runner,
@@ -112,7 +112,7 @@ def test_run_claude_extra_env_is_added_to_subprocess_env():
     runner, captured = _make_runner(stdout="")
     run_claude(
         prompt="hi",
-        model="claude-opus-4-7",
+        model="claude-opus-4-8",
         base_url="http://localhost",
         api_key="sk-abc",
         extra_env={"MAX_THINKING_TOKENS": "4096"},
@@ -146,7 +146,7 @@ def test_run_claude_inherits_only_allowlisted_os_environ(monkeypatch):
     runner, captured = _make_runner(stdout="")
     run_claude(
         prompt="hi",
-        model="claude-opus-4-7",
+        model="claude-opus-4-8",
         base_url="http://localhost",
         api_key="sk-abc",
         runner=runner,
@@ -178,7 +178,7 @@ def test_run_claude_uses_isolated_per_invocation_home(monkeypatch, tmp_path):
     runner, captured = _make_runner(stdout="")
     run_claude(
         prompt="hi",
-        model="claude-opus-4-7",
+        model="claude-opus-4-8",
         base_url="http://localhost",
         api_key="sk-abc",
         runner=runner,
@@ -209,7 +209,7 @@ def test_run_claude_isolated_home_is_distinct_per_invocation(monkeypatch):
     runner, captured = _make_runner(stdout="")
     run_claude(
         prompt="hi",
-        model="claude-opus-4-7",
+        model="claude-opus-4-8",
         base_url="http://localhost",
         api_key="sk-abc",
         runner=runner,
@@ -219,7 +219,7 @@ def test_run_claude_isolated_home_is_distinct_per_invocation(monkeypatch):
     runner, captured = _make_runner(stdout="")
     run_claude(
         prompt="hi",
-        model="claude-opus-4-7",
+        model="claude-opus-4-8",
         base_url="http://localhost",
         api_key="sk-abc",
         runner=runner,
@@ -243,7 +243,7 @@ def test_run_claude_isolated_home_cleaned_up_after_run(monkeypatch):
     runner, captured = _make_runner(stdout="")
     run_claude(
         prompt="hi",
-        model="claude-opus-4-7",
+        model="claude-opus-4-8",
         base_url="http://localhost",
         api_key="sk-abc",
         runner=runner,
@@ -275,7 +275,7 @@ def runner(cmd, env, capture_output, text, timeout, check, input=None):
     with pytest.raises(ClaudeCLIError):
         run_claude(
             prompt="hi",
-            model="claude-opus-4-7",
+            model="claude-opus-4-8",
             base_url="http://localhost",
             api_key="sk-abc",
             runner=runner,
@@ -294,7 +294,7 @@ def test_run_claude_extra_env_can_pass_through_otherwise_blocked_var(monkeypatch
     runner, captured = _make_runner(stdout="")
     run_claude(
         prompt="hi",
-        model="claude-opus-4-7",
+        model="claude-opus-4-8",
         base_url="http://localhost",
         api_key="sk-abc",
         extra_env={"ANTHROPIC_API_KEY": "from-arg"},

tests/e2e/claude_code/_driver_unit_tests/test_rate_limiter.py

@@ -60,11 +60,11 @@
     "model, expected",
     [
         ("claude-haiku-4-5", PROVIDER_ANTHROPIC),
-        ("claude-sonnet-4-6", PROVIDER_ANTHROPIC),
-        ("claude-opus-4-7", PROVIDER_ANTHROPIC),
+        ("claude-sonnet-5", PROVIDER_ANTHROPIC),
+        ("claude-opus-4-8", PROVIDER_ANTHROPIC),
         ("claude-haiku-4-5-azure", PROVIDER_AZURE),
-        ("claude-sonnet-4-6-azure", PROVIDER_AZURE),
-        ("claude-opus-4-7-vertex", PROVIDER_VERTEX_AI),
+        ("claude-sonnet-5-azure", PROVIDER_AZURE),
+        ("claude-opus-4-8-vertex", PROVIDER_VERTEX_AI),
         ("claude-haiku-4-5-bedrock-converse", PROVIDER_BEDROCK_CONVERSE),
         ("claude-haiku-4-5-bedrock-invoke", PROVIDER_BEDROCK_INVOKE),
     ],
@@ -87,7 +87,7 @@ def test_infer_provider_rejects_empty_string():
 def test_infer_provider_is_case_insensitive():
     """Aliases in the proxy config sometimes drift between cases; we
     should still route them to the right column."""
-    assert infer_provider("CLAUDE-OPUS-4-7-AZURE") == PROVIDER_AZURE
+    assert infer_provider("CLAUDE-OPUS-4-8-AZURE") == PROVIDER_AZURE
 
 
 # ---------------------------------------------------------------------------

tests/e2e/claude_code/basic_messaging_non_streaming/test_anthropic.py

@@ -11,8 +11,8 @@
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^      ^^^^^^^^^
                        feature_id                         provider
 
-Per the PRD, every cell exercises Claude Haiku 4.5, Sonnet 4.6, and Opus
-4.7; the cell only goes green if all three pass. The shared
+Per the PRD, every cell exercises Claude Haiku 4.5, Sonnet 5, and Opus
+4.8; the cell only goes green if all three pass. The shared
 `run_basic_messaging_cell` helper fans the three model runs out in
 parallel and reports one `compat_result.add(...)` entry per model so
 the matrix builder still sees three rows for this (feature, provider).
@@ -27,8 +27,8 @@
 # routing config; the driver only sends the alias.
 ANTHROPIC_MODELS = [
     "claude-haiku-4-5",
-    "claude-sonnet-4-6",
-    "claude-opus-4-7",
+    "claude-sonnet-5",
+    "claude-opus-4-8",
 ]
 
 

tests/e2e/claude_code/basic_messaging_non_streaming/test_azure.py

@@ -17,7 +17,7 @@
                        feature_id                          provider
 
 Per the PRD, every cell exercises Claude Haiku 4.5, Sonnet 4.6, and Opus
-4.7; the cell only goes green if all three pass. The shared
+4.8; the cell only goes green if all three pass. The shared
 `run_basic_messaging_cell` helper fans the three model runs out in
 parallel and reports one `compat_result.add(...)` entry per model so
 the matrix builder still sees three rows for this (feature, provider).
@@ -34,7 +34,7 @@
 AZURE_MODELS = [
     "claude-haiku-4-5-azure",
     "claude-sonnet-4-6-azure",
-    "claude-opus-4-7-azure",
+    "claude-opus-4-8-azure",
 ]
 
 

tests/e2e/claude_code/basic_messaging_non_streaming/test_bedrock_converse.py

@@ -11,8 +11,8 @@
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^      ^^^^^^^^^^^^^^^^
                        feature_id                          provider
 
-Per the PRD, every cell exercises Claude Haiku 4.5, Sonnet 4.6, and Opus
-4.7; the cell only goes green if all three pass. The shared
+Per the PRD, every cell exercises Claude Haiku 4.5, Sonnet 5, and Opus
+4.8; the cell only goes green if all three pass. The shared
 `run_basic_messaging_cell` helper fans the three model runs out in
 parallel and reports one `compat_result.add(...)` entry per model so
 the matrix builder still sees three rows for this (feature, provider).
@@ -28,8 +28,8 @@
 # strategy.
 BEDROCK_CONVERSE_MODELS = [
     "claude-haiku-4-5-bedrock-converse",
-    "claude-sonnet-4-6-bedrock-converse",
-    "claude-opus-4-7-bedrock-converse",
+    "claude-sonnet-5-bedrock-converse",
+    "claude-opus-4-8-bedrock-converse",
 ]
 
 

tests/e2e/claude_code/basic_messaging_non_streaming/test_bedrock_invoke.py

@@ -11,8 +11,8 @@
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^      ^^^^^^^^^^^^^^
                        feature_id                          provider
 
-Per the PRD, every cell exercises Claude Haiku 4.5, Sonnet 4.6, and Opus
-4.7; the cell only goes green if all three pass. The shared
+Per the PRD, every cell exercises Claude Haiku 4.5, Sonnet 5, and Opus
+4.8; the cell only goes green if all three pass. The shared
 `run_basic_messaging_cell` helper fans the three model runs out in
 parallel and reports one `compat_result.add(...)` entry per model so
 the matrix builder still sees three rows for this (feature, provider).
@@ -28,8 +28,8 @@
 # routing strategy.
 BEDROCK_INVOKE_MODELS = [
     "claude-haiku-4-5-bedrock-invoke",
-    "claude-sonnet-4-6-bedrock-invoke",
-    "claude-opus-4-7-bedrock-invoke",
+    "claude-sonnet-5-bedrock-invoke",
+    "claude-opus-4-8-bedrock-invoke",
 ]
 
 

tests/e2e/claude_code/basic_messaging_streaming/test_anthropic.py

@@ -29,8 +29,8 @@
 
 ANTHROPIC_MODELS = [
     "claude-haiku-4-5",
-    "claude-sonnet-4-6",
-    "claude-opus-4-7",
+    "claude-sonnet-5",
+    "claude-opus-4-8",
 ]
 
 

tests/e2e/claude_code/basic_messaging_streaming/test_azure.py

@@ -24,7 +24,7 @@
 AZURE_MODELS = [
     "claude-haiku-4-5-azure",
     "claude-sonnet-4-6-azure",
-    "claude-opus-4-7-azure",
+    "claude-opus-4-8-azure",
 ]
 
 

tests/e2e/claude_code/basic_messaging_streaming/test_bedrock_converse.py

@@ -19,8 +19,8 @@
 
 BEDROCK_CONVERSE_MODELS = [
     "claude-haiku-4-5-bedrock-converse",
-    "claude-sonnet-4-6-bedrock-converse",
-    "claude-opus-4-7-bedrock-converse",
+    "claude-sonnet-5-bedrock-converse",
+    "claude-opus-4-8-bedrock-converse",
 ]
 
 

tests/e2e/claude_code/basic_messaging_streaming/test_bedrock_invoke.py

@@ -19,8 +19,8 @@
 
 BEDROCK_INVOKE_MODELS = [
     "claude-haiku-4-5-bedrock-invoke",
-    "claude-sonnet-4-6-bedrock-invoke",
-    "claude-opus-4-7-bedrock-invoke",
+    "claude-sonnet-5-bedrock-invoke",
+    "claude-opus-4-8-bedrock-invoke",
 ]