{ "meta": { "benchmark_version": "2026.09.1", "generated_at": "2026-09-09T02:50:00+08:00", "period": "Last 30 Days (2026-08-10 to 2026-09-09)", "test_protocol": "HTTP/2 & HTTP/1.1 SSE Streaming (100-token prompt payload, 50-token output)", "total_synthetic_samples": 10000, "monitoring_locations": [ { "id": "hkg-01", "name": "Hong Kong BGP (CN2/CMI)", "ip_prefix": "103.140.*.*" }, { "id": "tyo-01", "name": "Tokyo AWS ap-northeast-1", "ip_prefix": "13.230.*.*" }, { "id": "sha-01", "name": "Shanghai Telecom CN2 Transit", "ip_prefix": "101.231.*.*" }, { "id": "pek-01", "name": "Beijing CERNET / China Mobile", "ip_prefix": "202.112.*.*" } ] }, "service_availability": { "target_sla": 0.9995, "measured_24h": 0.9999, "measured_7d": 0.9998, "measured_30d": 0.99965, "measured_90d": 0.99952, "total_downtime_minutes_30d": 15.2, "uptime_breakdown_by_upstream": { "openai_enterprise": 0.9996, "anthropic_direct": 0.9995, "google_vertex": 0.9998, "deepseek_weight": 0.9997 } }, "latency_ttft_ms": { "definition": "Time To First Token measured from request transmission to initial stream chunk arrival", "models": { "gemini-3.8-flash": { "tokyo_aws": { "p50": 112, "p90": 145, "p95": 168, "p99": 240, "sample_count": 2500 }, "hong_kong": { "p50": 165, "p90": 198, "p95": 224, "p99": 310, "sample_count": 2500 }, "shanghai": { "p50": 182, "p90": 235, "p95": 272, "p99": 380, "sample_count": 2500 }, "beijing": { "p50": 195, "p90": 250, "p95": 298, "p99": 415, "sample_count": 2500 } }, "claude-sonnet-5": { "tokyo_aws": { "p50": 148, "p90": 210, "p95": 255, "p99": 360, "sample_count": 2500 }, "hong_kong": { "p50": 192, "p90": 260, "p95": 305, "p99": 420, "sample_count": 2500 }, "shanghai": { "p50": 210, "p90": 285, "p95": 330, "p99": 465, "sample_count": 2500 }, "beijing": { "p50": 230, "p90": 310, "p95": 365, "p99": 510, "sample_count": 2500 } }, "gpt-6-astra": { "tokyo_aws": { "p50": 175, "p90": 240, "p95": 290, "p99": 410, "sample_count": 2500 }, "hong_kong": { "p50": 215, "p90": 295, "p95": 345, "p99": 485, "sample_count": 2500 }, "shanghai": { "p50": 235, "p90": 320, "p95": 375, "p99": 525, "sample_count": 2500 }, "beijing": { "p50": 255, "p90": 345, "p95": 410, "p99": 570, "sample_count": 2500 } }, "gpt-5.6-sol": { "tokyo_aws": { "p50": 160, "p90": 225, "p95": 270, "p99": 390, "sample_count": 2500 }, "hong_kong": { "p50": 200, "p90": 275, "p95": 320, "p99": 460, "sample_count": 2500 }, "shanghai": { "p50": 220, "p90": 300, "p95": 350, "p99": 490, "sample_count": 2500 }, "beijing": { "p50": 240, "p90": 325, "p95": 385, "p99": 540, "sample_count": 2500 } } } }, "modeloc_behavior_verification": { "test_method": "MODELOC 13-Probe Behavioral Protocol", "summary_rank": "High Fidelity (Score 96.8 / 100)", "verification_table": [ { "jizhi_model_id": "gpt-6-astra", "official_id": "gpt-6-astra", "provider": "OpenAI", "type": "upstream", "score": "96/100", "risk_level": "Low Risk", "probes_passed": "13/13", "last_probe_time": "2026-09-09T02:40:12+08:00", "result": "Passed" }, { "jizhi_model_id": "gpt-5.6-sol", "official_id": "gpt-5.6-sol", "provider": "OpenAI", "type": "upstream", "score": "95/100", "risk_level": "Low Risk", "probes_passed": "13/13", "last_probe_time": "2026-09-09T02:41:05+08:00", "result": "Passed" }, { "jizhi_model_id": "claude-sonnet-5", "official_id": "claude-sonnet-5", "provider": "Anthropic", "type": "upstream", "score": "98/100", "risk_level": "Low Risk", "probes_passed": "13/13", "last_probe_time": "2026-09-09T02:42:01+08:00", "result": "Passed" }, { "jizhi_model_id": "claude-opus-4-6", "official_id": "claude-opus-4-6", "provider": "Anthropic", "type": "upstream", "score": "97/100", "risk_level": "Low Risk", "probes_passed": "13/13", "last_probe_time": "2026-09-09T02:42:19+08:00", "result": "Passed" }, { "jizhi_model_id": "gemini-3.8-flash", "official_id": "gemini-3.8-flash", "provider": "Google", "type": "upstream", "score": "99/100", "risk_level": "Low Risk", "probes_passed": "13/13", "last_probe_time": "2026-09-09T02:43:40+08:00", "result": "Passed" } ] }, "rate_limits": { "standard_tier": { "requests_per_minute": 6000, "tokens_per_minute": 500000, "concurrent_connections": 100, "connection_timeout_seconds": 600 }, "enterprise_tier": { "requests_per_minute": 30000, "tokens_per_minute": 3000000, "concurrent_connections": 500, "connection_timeout_seconds": 1200 } } }