{
  "name": "TokenAAS Chinese AI model API benchmark baseline",
  "version": "2026-09-11",
  "license": "CC BY 4.0",
  "publisher": "TokenAAS",
  "url": "https://tokenaas.ai/guides/chinese-ai-model-api-benchmark",
  "scope": "Reproducible API performance measurements for Chinese AI model access through TokenAAS",
  "methodology": {
    "request_mode": "streaming chat",
    "sample_size_per_route": 5,
    "metrics": [
      "first_token_latency_p50_ms",
      "first_token_latency_p95_ms",
      "total_latency_p50_ms",
      "total_latency_p95_ms",
      "success_rate_percent",
      "timeout_rate_percent"
    ],
    "limitations": [
      "The initial dataset measures routing paths for one model, not quality differences between model families.",
      "Five requests per route are sufficient only for an initial reproducible baseline.",
      "Results can change with region, time window, prompt, model version, upstream capacity, and network conditions."
    ]
  },
  "published_results": [
    {
      "test_date": "2026-08-31",
      "model": "deepseek-v4-flash",
      "route": "direct",
      "samples": 5,
      "first_token_latency_p50_ms": 1240.62,
      "first_token_latency_p95_ms": 1448.30,
      "total_latency_p50_ms": 1268.71,
      "total_latency_p95_ms": 1492.09,
      "success_rate_percent": 100,
      "timeout_rate_percent": 0
    },
    {
      "test_date": "2026-08-31",
      "model": "deepseek-v4-flash",
      "route": "authenticated_regional_relay",
      "samples": 5,
      "first_token_latency_p50_ms": 1240.14,
      "first_token_latency_p95_ms": 1328.71,
      "total_latency_p50_ms": 1300.00,
      "total_latency_p95_ms": 1395.36,
      "success_rate_percent": 100,
      "timeout_rate_percent": 0
    }
  ],
  "planned_model_matrix": [
    { "family": "DeepSeek", "candidate_ids": ["deepseek-v4-flash", "deepseek-v4-pro"], "results_status": "partial" },
    { "family": "Qwen", "candidate_ids": ["qwen3.7-flash"], "results_status": "pending" },
    { "family": "GLM", "candidate_ids": ["glm-5.2", "glm-5.3"], "results_status": "pending" },
    { "family": "Kimi", "candidate_ids": ["kimi-k2.6"], "results_status": "pending" }
  ]
}
