{ "context": { "cap_tokens": 31680, "max_model_len": 32768, "targets": [ { "attempted_context_tokens": 8192, "chars_per_token": 3.0, "clamped_tokens": 8192, "e2e_ms": 852.0760536193848, "estimated_prompt_tokens": 8192, "output_tokens": 52, "status": "passed", "target_tokens": 8192, "ttft_ms": 510.2362632751465, "usage": { "completion_tokens": 53, "prompt_tokens": 6521, "total_tokens": 6574 } }, { "attempted_context_tokens": 16384, "chars_per_token": 3.769, "clamped_tokens": 16384, "e2e_ms": 823.3175277709961, "estimated_prompt_tokens": 16383, "output_tokens": 47, "status": "passed", "target_tokens": 16384, "ttft_ms": 480.39841651916504, "usage": { "completion_tokens": 48, "prompt_tokens": 16360, "total_tokens": 16408 } }, { "attempted_context_tokens": 30000, "chars_per_token": 3.774, "clamped_tokens": 30000, "e2e_ms": 1522.4061012268066, "estimated_prompt_tokens": 29999, "output_tokens": 49, "status": "passed", "target_tokens": 30000, "ttft_ms": 1127.856731414795, "usage": { "completion_tokens": 51, "prompt_tokens": 32604, "total_tokens": 32655 } } ] }, "evaluation_protocol": { "budget_semantics": "visible_answer_tokens plus reasoning_headroom_tokens are sent as one max_tokens cap; the endpoint does not hard-partition the channels", "minimum_comparable_repetitions": 3, "minimum_pass_rate": 1.0, "reasoning_headroom_tokens": 0, "records_finish_reason": true, "records_full_visible_answer": true, "records_reasoning_channel_metadata": true, "repetitions": 3, "version": 3, "visible_answer_tokens": 512 }, "failures": [], "identity": { "base_url": "http://127.0.0.1:30003/v1", "candidate_id": "legacy-e4b-fp8-dynamic", "config_id": "fast-control-32k-old-template", "engine": "vllm-nightly-a65f93f", "gpu": "dark-fast-rtx5090", "model": "gemma4-e4b-it", "started_at": "2026-07-16T16:50:48Z" }, "intelligence": { "checks": [ { "attempt_count": 3, "attempts": [ { "attempt": 1, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "```diff\n--- a/app.py\n+++ b/app.py\n@@ -1,2 +1,2 @@\n-timeout = 30\n+timeout = 45\n retries = 2\n```", "content_excerpt": "```diff\n--- a/app.py\n+++ b/app.py\n@@ -1,2 +1,2 @@\n-timeout = 30\n+timeout = 45\n retries = 2\n```", "failure_class": null, "finish_reason": "stop", "latency_ms": 294.59166526794434, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "text_checks": [ { "name": "diff_shape", "passed": true }, { "name": "removes_old_timeout", "passed": true }, { "name": "adds_new_timeout", "passed": true } ], "usage": { "completion_tokens": 50, "prompt_tokens": 50, "prompt_tokens_details": null, "total_tokens": 100 } }, { "attempt": 2, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "```diff\n--- a/app.py\n+++ b/app.py\n@@ -1,2 +1,2 @@\n-timeout = 30\n+timeout = 45\n retries = 2\n```", "content_excerpt": "```diff\n--- a/app.py\n+++ b/app.py\n@@ -1,2 +1,2 @@\n-timeout = 30\n+timeout = 45\n retries = 2\n```", "failure_class": null, "finish_reason": "stop", "latency_ms": 302.4313449859619, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "text_checks": [ { "name": "diff_shape", "passed": true }, { "name": "removes_old_timeout", "passed": true }, { "name": "adds_new_timeout", "passed": true } ], "usage": { "completion_tokens": 50, "prompt_tokens": 50, "prompt_tokens_details": null, "total_tokens": 100 } }, { "attempt": 3, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "```diff\n--- a/app.py\n+++ b/app.py\n@@ -1,2 +1,2 @@\n-timeout = 30\n+timeout = 45\n retries = 2\n```", "content_excerpt": "```diff\n--- a/app.py\n+++ b/app.py\n@@ -1,2 +1,2 @@\n-timeout = 30\n+timeout = 45\n retries = 2\n```", "failure_class": null, "finish_reason": "stop", "latency_ms": 299.35550689697266, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "text_checks": [ { "name": "diff_shape", "passed": true }, { "name": "removes_old_timeout", "passed": true }, { "name": "adds_new_timeout", "passed": true } ], "usage": { "completion_tokens": 50, "prompt_tokens": 50, "prompt_tokens_details": null, "total_tokens": 100 } } ], "id": "unified_diff_timeout_edit", "pass_count": 3, "pass_rate": 1.0, "required_pass_rate": 1.0, "status": "passed", "validator": "deterministic_text_checks" }, { "attempt_count": 3, "attempts": [ { "attempt": 1, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "The total processing time (3000 ms) exceeds the 2500 ms timeout, so optimize the LLM response time or increase the timeout limit.", "content_excerpt": "The total processing time (3000 ms) exceeds the 2500 ms timeout, so optimize the LLM response time or increase the timeout limit.", "failure_class": null, "finish_reason": "stop", "latency_ms": 206.38585090637207, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "text_checks": [ { "name": "identifies_budget_overrun", "passed": true }, { "name": "offers_latency_fix", "passed": true } ], "usage": { "completion_tokens": 35, "prompt_tokens": 74, "prompt_tokens_details": null, "total_tokens": 109 } }, { "attempt": 2, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "The total processing time (3000 ms) exceeds the 2500 ms timeout, so optimize the LLM response time or increase the timeout limit.", "content_excerpt": "The total processing time (3000 ms) exceeds the 2500 ms timeout, so optimize the LLM response time or increase the timeout limit.", "failure_class": null, "finish_reason": "stop", "latency_ms": 214.34617042541504, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "text_checks": [ { "name": "identifies_budget_overrun", "passed": true }, { "name": "offers_latency_fix", "passed": true } ], "usage": { "completion_tokens": 35, "prompt_tokens": 74, "prompt_tokens_details": null, "total_tokens": 109 } }, { "attempt": 3, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "The total processing time (3000 ms) exceeds the 2500 ms timeout, so optimize the LLM response time or increase the timeout limit.", "content_excerpt": "The total processing time (3000 ms) exceeds the 2500 ms timeout, so optimize the LLM response time or increase the timeout limit.", "failure_class": null, "finish_reason": "stop", "latency_ms": 217.07439422607422, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "text_checks": [ { "name": "identifies_budget_overrun", "passed": true }, { "name": "offers_latency_fix", "passed": true } ], "usage": { "completion_tokens": 35, "prompt_tokens": 74, "prompt_tokens_details": null, "total_tokens": 109 } } ], "id": "parallel_timeout_triage", "pass_count": 3, "pass_rate": 1.0, "required_pass_rate": 1.0, "status": "passed", "validator": "deterministic_text_checks" } ], "status": "passed" }, "run_id": "fast-bakeoff-20260716T165048Z", "schema": "anvil-serving.fast-tier-bakeoff/v1", "score_inputs": { "e2e_p50_ms": 852.0760536193848, "intelligence_pass_rate": 1.0, "operational_fit_notes": [ "endpoint was already loaded; benchmark did not start or stop serves" ], "session_recall_passed": true, "thinking_mode": "disabled", "tool_call_passed": true, "ttft_p50_ms": 510.2362632751465, "usable_context_tokens": 30000, "voice_latency_ms": null }, "selection": { "context_targets": [ 8192, 16384, 30000 ], "endpoint_already_loaded": true, "requests_per_context": 1, "suites": [ "chat", "context", "tool", "session", "intelligence" ] }, "session": { "checks": [ { "attempt_count": 3, "attempts": [ { "attempt": 1, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "RIVER-918", "content_excerpt": "RIVER-918", "expected": "RIVER-918", "failure_class": null, "finish_reason": "stop", "latency_ms": 60.98198890686035, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "usage": { "completion_tokens": 7, "prompt_tokens": 48, "prompt_tokens_details": null, "total_tokens": 55 } }, { "attempt": 2, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "RIVER-918", "content_excerpt": "RIVER-918", "expected": "RIVER-918", "failure_class": null, "finish_reason": "stop", "latency_ms": 55.8466911315918, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "usage": { "completion_tokens": 7, "prompt_tokens": 48, "prompt_tokens_details": null, "total_tokens": 55 } }, { "attempt": 3, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "RIVER-918", "content_excerpt": "RIVER-918", "expected": "RIVER-918", "failure_class": null, "finish_reason": "stop", "latency_ms": 56.30636215209961, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "usage": { "completion_tokens": 7, "prompt_tokens": 48, "prompt_tokens_details": null, "total_tokens": 55 } } ], "name": "single_request_multiturn_recall", "pass_count": 3, "pass_rate": 1.0, "required_pass_rate": 1.0, "status": "passed" } ], "status": "passed" }, "source_recipe": { "ref": "examples/primary-node/docker-compose.yml#fast", "serve_command": "anvil-serving serves up fast --manifest examples/primary-node/serves.toml --confirm" }, "suites": {}, "thinking": { "chat_template_kwargs": { "enable_thinking": false }, "control_evidence": "C:\\Users\\operator\\ai-code\\anvil-serving\\docs\\findings\\2026-07-16-gemma4-chat-template-bakeoff-evidence\\fast-control-legacy-thinking-control.json", "control_evidence_sha256": "a93914fd916d9bff2ac776e189c6697857dc85da648a296702fda4b6c45f7640", "control_mechanism": "chat_template_kwargs", "control_requested": { "enable_thinking": false }, "control_status": "verified", "mode": "disabled", "reasoning_effort": null, "unsupported": false }, "timing": { "chat": { "e2e_p50_ms": 852.0760536193848, "e2e_p95_ms": 1522.4061012268066, "output_tokens": 148, "ttft_p50_ms": 510.2362632751465, "ttft_p95_ms": 1127.856731414795 }, "wall_ms": 5348.1316566467285 }, "tool": { "checks": [ { "attempt_count": 3, "attempts": [ { "arguments": { "zip": "98101" }, "attempt": 1, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "", "content_excerpt": "", "failure_class": null, "finish_reason": "tool_calls", "latency_ms": 141.4637565612793, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "tool_call_count": 1, "usage": { "completion_tokens": 21, "prompt_tokens": 79, "prompt_tokens_details": null, "total_tokens": 100 }, "valid_tool_call_count": 1, "validation_errors": [] }, { "arguments": { "zip": "98101" }, "attempt": 2, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "", "content_excerpt": "", "failure_class": null, "finish_reason": "tool_calls", "latency_ms": 131.32357597351074, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "tool_call_count": 1, "usage": { "completion_tokens": 21, "prompt_tokens": 79, "prompt_tokens_details": null, "total_tokens": 100 }, "valid_tool_call_count": 1, "validation_errors": [] }, { "arguments": { "zip": "98101" }, "attempt": 3, "budget": { "legacy_max_tokens_as_visible": false, "max_completion_tokens": 512, "reasoning_headroom_tokens": 0, "visible_answer_tokens": 512 }, "content": "", "content_excerpt": "", "failure_class": null, "finish_reason": "tool_calls", "latency_ms": 136.65437698364258, "reasoning_chars": 0, "reasoning_excerpt": "", "reasoning_field": null, "reasoning_tokens": null, "status": "passed", "tool_call_count": 1, "usage": { "completion_tokens": 21, "prompt_tokens": 79, "prompt_tokens_details": null, "total_tokens": 100 }, "valid_tool_call_count": 1, "validation_errors": [] } ], "expected_arguments": { "zip": "98101" }, "expected_function": "record_weather_zip", "name": "openai_tool_call_smoke", "pass_count": 3, "pass_rate": 1.0, "required_pass_rate": 1.0, "status": "passed" } ], "status": "passed" }, "voice": { "llm_latency_ms": null, "status": "not_run", "stt_latency_ms": null, "total_turn_latency_ms": null, "tts_latency_ms": null } }