{ "base_url": "http://127.0.0.1:8080/v1", "clock_domain_id": null, "completed": 12, "concurrency": 1, "configuration_fingerprint": null, "context_distribution": { "4096": 12 }, "context_seed": 73, "context_tokens": 4096, "controlled_output_policy": "strict", "engine": "ninfer", "failed": 0, "failures": [], "finished_at": "2026-09-20T01:14:12Z", "finished_at_unix_ns": 1789866852233599900, "gpu": null, "manifest": null, "max_context_tokens": 163840, "max_tokens": 512, "measurement_protocol": "capacity-v3", "metric_population": { "aggregate_throughput_eligible": true, "attempted_requests": 12, "excluded_requests": 0, "failure_policy": "failed or contract-invalid responses retain sanitized timing and metadata but are excluded from performance distributions", "p99_interpretation": "descriptive-nearest-rank-only", "p99_interpretation_by_metric": { "decode_tok_s": "descriptive-nearest-rank-only", "e2e": "descriptive-nearest-rank-only", "effective_prefill_tok_s": "descriptive-nearest-rank-only", "tpot": "descriptive-nearest-rank-only", "ttft": "descriptive-nearest-rank-only" }, "p99_minimum_recommended_samples": 100, "p99_samples_by_metric": { "decode_tok_s": 12, "e2e": 12, "effective_prefill_tok_s": 12, "tpot": 12, "ttft": 12 }, "performance_eligible_requests": 12, "successful_requests": 12 }, "metrics": { "completion_tokens_max": 46.0, "completion_tokens_mean": 46.0, "completion_tokens_mean_ci95_high": 46.0, "completion_tokens_mean_ci95_low": 46.0, "completion_tokens_min": 46.0, "completion_tokens_p25": 46.0, "completion_tokens_p50": 46.0, "completion_tokens_p75": 46.0, "completion_tokens_p90": 46.0, "completion_tokens_p95": 46.0, "completion_tokens_p99": 46.0, "completion_tokens_samples": 12, "completion_tokens_stddev": 0.0, "content_chunks": 156, "content_chunks_s": null, "controlled_output_exact_adherence": 12, "controlled_output_exact_adherence_rate": 1.0, "controlled_output_samples": 12, "controlled_output_unobservable": 0, "decode_tok_s_max": 189.54487331488414, "decode_tok_s_mean": 179.77670707526696, "decode_tok_s_mean_ci95_high": 183.8478872403258, "decode_tok_s_mean_ci95_low": 175.7055269102081, "decode_tok_s_min": 174.59631391569172, "decode_tok_s_p25": 174.82279765018768, "decode_tok_s_p50": 174.98547623747788, "decode_tok_s_p75": 189.46275517357248, "decode_tok_s_p90": 189.53369661233586, "decode_tok_s_p95": 189.54487331488414, "decode_tok_s_p99": 189.54487331488414, "decode_tok_s_samples": 12, "decode_tok_s_stddev": 7.195398870049588, "e2e_max_ms": 889.1625000396743, "e2e_mean_ci95_high_ms": 788.7631565877894, "e2e_mean_ci95_low_ms": 743.4958433924207, "e2e_mean_ms": 766.129499990105, "e2e_min_ms": 740.6920999055728, "e2e_p25_ms": 742.0409000478685, "e2e_p50_ms": 761.8653998943046, "e2e_p75_ms": 762.5608999514952, "e2e_p90_ms": 763.3569999597967, "e2e_p95_ms": 889.1625000396743, "e2e_p99_ms": 889.1625000396743, "e2e_samples": 12, "e2e_stddev_ms": 40.002697130873294, "effective_prefill_tok_s_max": 7183.093565681958, "effective_prefill_tok_s_mean": 7048.409096224807, "effective_prefill_tok_s_mean_ci95_high": 7287.313403078331, "effective_prefill_tok_s_mean_ci95_low": 6809.504789371284, "effective_prefill_tok_s_min": 5707.820314328614, "effective_prefill_tok_s_p25": 7163.14204646255, "effective_prefill_tok_s_p50": 7166.749136424184, "effective_prefill_tok_s_p75": 7172.465812843637, "effective_prefill_tok_s_p90": 7180.727522434564, "effective_prefill_tok_s_p95": 7183.093565681958, "effective_prefill_tok_s_p99": 7183.093565681958, "effective_prefill_tok_s_samples": 12, "effective_prefill_tok_s_stddev": 422.23918124217226, "generation_max_ms": 257.73740001022816, "generation_mean_ci95_high_ms": 256.1961044434289, "generation_mean_ci95_low_ms": 245.14172885168173, "generation_mean_ms": 250.6689166475553, "generation_min_ms": 237.41079994942993, "generation_p25_ms": 237.43459989782423, "generation_p50_ms": 257.12940003722906, "generation_p75_ms": 257.35019997227937, "generation_p90_ms": 257.418500026688, "generation_p95_ms": 257.73740001022816, "generation_p99_ms": 257.73740001022816, "generation_samples": 12, "generation_stddev_ms": 9.768744985130285, "mean_inter_token_latency_ms_max": 5.72749777800507, "mean_inter_token_latency_ms_mean": 5.570420369945674, "mean_inter_token_latency_ms_mean_ci95_high": 5.693246765409531, "mean_inter_token_latency_ms_mean_ci95_low": 5.4475939744818165, "mean_inter_token_latency_ms_min": 5.2757955544317765, "mean_inter_token_latency_ms_p25": 5.2763244421738715, "mean_inter_token_latency_ms_p50": 5.713986667493979, "mean_inter_token_latency_ms_p75": 5.718893332717319, "mean_inter_token_latency_ms_p90": 5.720411111704177, "mean_inter_token_latency_ms_p95": 5.72749777800507, "mean_inter_token_latency_ms_p99": 5.72749777800507, "mean_inter_token_latency_ms_samples": 12, "mean_inter_token_latency_ms_stddev": 0.21708322189178406, "mean_time_per_output_token_max_ms": 5.72749777800507, "mean_time_per_output_token_mean_ci95_high_ms": 5.693246765409531, "mean_time_per_output_token_mean_ci95_low_ms": 5.4475939744818165, "mean_time_per_output_token_mean_ms": 5.570420369945674, "mean_time_per_output_token_min_ms": 5.2757955544317765, "mean_time_per_output_token_p25_ms": 5.2763244421738715, "mean_time_per_output_token_p50_ms": 5.713986667493979, "mean_time_per_output_token_p75_ms": 5.718893332717319, "mean_time_per_output_token_p90_ms": 5.720411111704177, "mean_time_per_output_token_p95_ms": 5.72749777800507, "mean_time_per_output_token_p99_ms": 5.72749777800507, "mean_time_per_output_token_samples": 12, "mean_time_per_output_token_stddev_ms": 0.21708322189178406, "output_token_sources": { "usage": 12 }, "output_tokens": 552, "prefix_cache_hit_avg": null, "prompt_token_samples": 12, "prompt_tokens": 43427, "prompt_tokens_max": 3628.0, "prompt_tokens_mean": 3618.9166666666665, "prompt_tokens_mean_ci95_high": 3622.498846325912, "prompt_tokens_mean_ci95_low": 3615.334487007421, "prompt_tokens_min": 3610.0, "prompt_tokens_p25": 3612.0, "prompt_tokens_p50": 3618.0, "prompt_tokens_p75": 3624.0, "prompt_tokens_p90": 3628.0, "prompt_tokens_p95": 3628.0, "prompt_tokens_p99": 3628.0, "prompt_tokens_samples": 12, "prompt_tokens_stddev": 6.331139971074193, "reasoning_chunks": 0, "request_canary_passed": 12, "request_canary_samples": 12, "throughput_performance_eligible": true, "throughput_tok_s": 60.0201478505868, "time_to_first_output_max_ms": 632.4656000360847, "time_to_first_output_mean_ci95_high_ms": 536.312465262299, "time_to_first_output_mean_ci95_low_ms": 494.6087014228004, "time_to_first_output_mean_ms": 515.4605833425497, "time_to_first_output_min_ms": 503.2575000077486, "time_to_first_output_p25_ms": 504.5272000133991, "time_to_first_output_p50_ms": 504.8234999412671, "time_to_first_output_p75_ms": 505.09720004629344, "time_to_first_output_p90_ms": 506.0067999875173, "time_to_first_output_p95_ms": 632.4656000360847, "time_to_first_output_p99_ms": 632.4656000360847, "time_to_first_output_samples": 12, "time_to_first_output_stddev_ms": 36.85359073309461, "tpot_max_ms": 5.72749777800507, "tpot_mean_ci95_high_ms": 5.693246765409531, "tpot_mean_ci95_low_ms": 5.4475939744818165, "tpot_mean_ms": 5.570420369945674, "tpot_min_ms": 5.2757955544317765, "tpot_p25_ms": 5.2763244421738715, "tpot_p50_ms": 5.713986667493979, "tpot_p75_ms": 5.718893332717319, "tpot_p90_ms": 5.720411111704177, "tpot_p95_ms": 5.72749777800507, "tpot_p99_ms": 5.72749777800507, "tpot_samples": 12, "tpot_stddev_ms": 0.21708322189178406, "ttft_max_ms": 632.4665000429377, "ttft_mean_ci95_high_ms": 536.3132068015138, "ttft_mean_ci95_low_ms": 494.609376541637, "ttft_mean_ms": 515.4612916715754, "ttft_min_ms": 503.2581998966634, "ttft_p25_ms": 504.52790001872927, "ttft_p50_ms": 504.82389994431287, "ttft_p75_ms": 505.0977000501007, "ttft_p90_ms": 506.00729999132454, "ttft_p95_ms": 632.4665000429377, "ttft_p99_ms": 632.4665000429377, "ttft_samples": 12, "ttft_stddev_ms": 36.85364942874233 }, "model": "qwen38-huihui-ninfer-schemafix-mtp3-160k", "performance_eligible": true, "prompt_cache_mode": "unique", "request_canaries": true, "request_timings": [ { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 175.304025873981, "e2e_ms": 889.1625000396743, "effective_prefill_tok_s": 5707.820314328614, "generation_ms": 256.6969000035897, "mean_inter_token_latency_ms": 5.704375555635327, "mean_time_per_output_token_ms": 5.704375555635327, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3610, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00000", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 0, "time_to_first_output_ms": 632.4656000360847, "tpot_ms": 5.704375555635327, "ttft_ms": 632.4665000429377, "visible_content_truncated": false, "visible_generation_ms": 256.69599999673665 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 189.52587373266135, "e2e_ms": 740.6920999055728, "effective_prefill_tok_s": 7177.240279468038, "generation_ms": 237.43459989782423, "mean_inter_token_latency_ms": 5.2763244421738715, "mean_time_per_output_token_ms": 5.2763244421738715, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3612, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00001", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 1, "time_to_first_output_ms": 503.2575000077486, "tpot_ms": 5.2763244421738715, "ttft_ms": 503.2581998966634, "visible_content_truncated": false, "visible_generation_ms": 237.4339000089094 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 189.53369661233586, "e2e_ms": 741.9886999996379, "effective_prefill_tok_s": 7158.657207878595, "generation_ms": 237.42479993961751, "mean_inter_token_latency_ms": 5.276106665324834, "mean_time_per_output_token_ms": 5.276106665324834, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3612, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00002", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 2, "time_to_first_output_ms": 504.5639000600204, "tpot_ms": 5.276106665324834, "ttft_ms": 504.5653000706807, "visible_content_truncated": false, "visible_generation_ms": 237.42339992895722 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 189.46275517357248, "e2e_ms": 742.0409000478685, "effective_prefill_tok_s": 7163.14204646255, "generation_ms": 237.51370003446937, "mean_inter_token_latency_ms": 5.278082222988208, "mean_time_per_output_token_ms": 5.278082222988208, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3614, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00003", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 3, "time_to_first_output_ms": 504.5272000133991, "tpot_ms": 5.278082222988208, "ttft_ms": 504.52790001872927, "visible_content_truncated": false, "visible_generation_ms": 237.51300002913922 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 174.98547623747788, "e2e_ms": 761.8653998943046, "effective_prefill_tok_s": 7164.635234511625, "generation_ms": 257.16419995296746, "mean_inter_token_latency_ms": 5.7147599989548326, "mean_time_per_output_token_ms": 5.7147599989548326, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3616, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00004", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 4, "time_to_first_output_ms": 504.7011999413371, "tpot_ms": 5.7147599989548326, "ttft_ms": 504.7017999459058, "visible_content_truncated": false, "visible_generation_ms": 257.16359994839877 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 174.81261057513194, "e2e_ms": 762.2499000281096, "effective_prefill_tok_s": 7166.749136424184, "generation_ms": 257.418500026688, "mean_inter_token_latency_ms": 5.720411111704177, "mean_time_per_output_token_ms": 5.720411111704177, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3618, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00005", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 5, "time_to_first_output_ms": 504.83140000142157, "tpot_ms": 5.720411111704177, "ttft_ms": 504.8323000082746, "visible_content_truncated": false, "visible_generation_ms": 257.41760001983494 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 189.54487331488414, "e2e_ms": 742.5079999957234, "effective_prefill_tok_s": 7164.9575560274525, "generation_ms": 237.41079994942993, "mean_inter_token_latency_ms": 5.2757955544317765, "mean_time_per_output_token_ms": 5.2757955544317765, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3619, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00006", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 6, "time_to_first_output_ms": 505.09720004629344, "tpot_ms": 5.2757955544317765, "ttft_ms": 505.0977000501007, "visible_content_truncated": false, "visible_generation_ms": 237.41029994562268 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 175.00915878730544, "e2e_ms": 761.9767000433058, "effective_prefill_tok_s": 7172.465812843637, "generation_ms": 257.12940003722906, "mean_inter_token_latency_ms": 5.713986667493979, "mean_time_per_output_token_ms": 5.713986667493979, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3621, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00007", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 7, "time_to_first_output_ms": 504.8473000060767, "tpot_ms": 5.713986667493979, "ttft_ms": 504.84800001140684, "visible_content_truncated": false, "visible_generation_ms": 257.1287000318989 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 174.86389765579068, "e2e_ms": 761.8609999772161, "effective_prefill_tok_s": 7183.093565681958, "generation_ms": 257.342999917455, "mean_inter_token_latency_ms": 5.718733331499, "mean_time_per_output_token_ms": 5.718733331499, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3624, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00008", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 8, "time_to_first_output_ms": 504.5180000597611, "tpot_ms": 5.718733331499, "ttft_ms": 504.51870006509125, "visible_content_truncated": false, "visible_generation_ms": 257.3422999121249 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 174.59631391569172, "e2e_ms": 762.5608999514952, "effective_prefill_tok_s": 7180.727522434564, "generation_ms": 257.73740001022816, "mean_inter_token_latency_ms": 5.72749777800507, "mean_time_per_output_token_ms": 5.72749777800507, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3625, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00009", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 9, "time_to_first_output_ms": 504.8234999412671, "tpot_ms": 5.72749777800507, "ttft_ms": 504.82389994431287, "visible_content_truncated": false, "visible_generation_ms": 257.73700000718236 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 174.8590053741835, "e2e_ms": 763.3569999597967, "effective_prefill_tok_s": 7169.864120579999, "generation_ms": 257.35019997227937, "mean_inter_token_latency_ms": 5.718893332717319, "mean_time_per_output_token_ms": 5.718893332717319, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3628, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00010", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 10, "time_to_first_output_ms": 506.0067999875173, "tpot_ms": 5.718893332717319, "ttft_ms": 506.00729999132454, "visible_content_truncated": false, "visible_generation_ms": 257.3496999684721 }, { "content_chunks": 13, "controlled_output": { "capture_complete": true, "exact_adherence": true, "observed_code_words": 32, "observed_extra_words": 0, "passed": true, "policy": "strict", "requested_words": 32 }, "decode_tok_s": 174.82279765018768, "e2e_ms": 763.2909000385553, "effective_prefill_tok_s": 7171.556358056468, "generation_ms": 257.4035000288859, "mean_inter_token_latency_ms": 5.720077778419687, "mean_time_per_output_token_ms": 5.720077778419687, "output_token_source": "usage", "output_tokens": 46, "planned_context_tokens": 4096, "prompt_tokens": 3628, "reasoning_chunks": 0, "request_canary": { "capture_complete": true, "foreign_marker_count": 0, "marker": "ANVIL_REQ_73_00011", "marker_at_start": true, "observed_marker_count": 1, "passed": true }, "request_index": 11, "time_to_first_output_ms": 505.88740000966936, "tpot_ms": 5.720077778419687, "ttft_ms": 505.8879000134766, "visible_content_truncated": false, "visible_generation_ms": 257.40300002507865 } ], "requests": 12, "response_words": 32, "responses_received": 12, "run_id": "benchmark-20260920T011403Z", "sampling": { "temperature": { "effective_request": 0.0, "requested": null, "sent": true }, "top_p": { "effective_request": null, "requested": null, "sent": false } }, "schema": "anvil-serving.benchmark/v1", "serve_flags": { "no_thinking": false, "reasoning_effort": null, "shared_prefix_burst": false, "thinking_mode": "unsupported" }, "shared_prefix_tokens": 0, "source_recipe": "REDACTED_LOCAL_PATH/.anvil-serving\\evidence\\2026-09-19-qwen38-context-envelope\\serve-recipes.160k.toml#lyf/Qwen3.8-27B-Huihui-NInfer-NVFP4-MTP3-160K-Baked", "started_at": "2026-09-20T01:14:03Z", "started_at_unix_ns": 1789866843036826700, "tier": null, "timestamp_precision": "nanosecond-unix-fields; ISO labels are second precision", "timing_methodology": { "artifact_timestamps": "time.time_ns Unix timestamps; independent artifacts may align them only when the operator supplied the same explicit clock_domain_id", "clock": "client time.perf_counter", "controlled_output": "--response-words adherence is recorded; observe preserves historical behavior while strict rejects non-exact or truncated completions; visible response capture is capped at 8192 characters", "decode": "usage completion tokens after the first token divided by client-observed generation time; completion tokens may include reasoning tokens", "distributions": "request-level mean, sample standard deviation, normal-approximation 95% confidence interval for the mean, min, p25, p50, p75, p90, p95, p99, and max using the nearest-rank percentile method", "effective_prefill": "usage.prompt_tokens divided by client-observed time to first output; includes queueing, scheduling, prefill, and first-token work", "generation": "client-observed E2E minus time to first output", "mean_inter_token_latency": "legacy key for a TPOT proxy: client-observed generation time divided by completion tokens after the first token; not raw per-token ITL", "p99": "numeric nearest-rank p99 is always retained for compatibility; it is descriptive only below 100 eligible samples", "request_canaries": "when enabled, each response must begin with its own synthetic marker and contain no other ANVIL_REQ marker; bounded/truncated captures fail closed and are excluded from performance metrics", "time_to_first_output": "request start through the first non-empty streamed reasoning or content delta", "tpot": "client-observed mean time per output token after the first output; identical to the legacy mean_inter_token_latency key", "ttft": "request start through the first non-empty streamed content delta" }, "wall_clock_ms": 9196.911699953489 }