API

AI 模型基准测试数据的编程接口。提供免费层级。

Free

100 requests/day

所有公开端点
JSON 响应
须包含归属链接

Requires "Data by BenchGecko" attribution link

Pro

$29/mo

10,000 requests/day

所有公开端点
CSV + JSON 导出
历史数据访问
无需归属标注

Enterprise

Custom

Unlimited

包含 Pro 全部功能
Webhook 通知
自定义基准测试
优先支持

免费层级归属要求

免费 API 访问要求在展示我们数据的页面上显示"Data by BenchGecko"链接。

<a href="https://benchgecko.ai">Data by BenchGecko</a>

端点

GET/api/v1/models

列出所有模型的分数和定价

参数

providerstringFilter by provider slug (e.g. "anthropic")

open_sourcebooleanFilter open source models only

sortstringSort by: avg_score, pricing_input, release_date, context_window

limitnumberResults per page (default 50, max 200)

响应

{
  "data": [
    {
      "slug": "claude-opus-4",
      "name": "Claude Opus 4",
      "provider": "Anthropic",
      "avg_score": 84.2,
      "scores": {
        "mmlu-pro": 85.7,
        "gpqa-diamond": 74.9,
        "humaneval-plus": 91.2
      },
      "pricing": { "input": 15.00, "output": 75.00 },
      "context_window": 200000,
      "release_date": "2026-03-15"
    }
  ],
  "meta": { "total": 14, "page": 1 }
}

GET/api/v1/models/:slug

获取单个模型的详细数据

响应

{
  "slug": "claude-opus-4",
  "name": "Claude Opus 4",
  "provider": { "name": "Anthropic", "slug": "anthropic" },
  "scores": [
    { "benchmark": "MMLU-Pro", "score": 85.7, "category": "knowledge" },
    { "benchmark": "GPQA Diamond", "score": 74.9, "category": "reasoning" }
  ],
  "pricing": { "input": 15.00, "output": 75.00 },
  "context_window": 200000,
  "max_output_tokens": 32000,
  "is_open_source": false,
  "release_date": "2026-03-15"
}

GET/api/v1/benchmarks

列出所有基准测试及最高分模型

参数

categorystringFilter by category: coding, reasoning, math, knowledge

响应

{
  "data": [
    {
      "slug": "swe-bench-verified",
      "name": "SWE-bench Verified",
      "category": "coding",
      "top_models": [
        { "name": "Claude Opus 4", "score": 72.5 },
        { "name": "Claude Sonnet 4", "score": 70.3 }
      ]
    }
  ]
}

GET/api/v1/compare

两个或多个模型的正面对比

参数

modelsstringComma-separated model slugs (e.g. "claude-opus-4,gpt-4-1,o3")

响应

{
  "models": ["claude-opus-4", "gpt-4-1", "o3"],
  "benchmarks": [
    {
      "name": "MMLU-Pro",
      "scores": { "claude-opus-4": 85.7, "gpt-4-1": 83.1, "o3": 87.2 }
    }
  ]
}

SDKs, Datasets & Open Source

Official client libraries for 11 languages, open datasets on Harvard Dataverse and Hugging Face, and full source code.

Explore the Developer Hub →

速率限制

Free: 100 requests/day, 10 requests/minute

Pro: 10,000 requests/day, 100 requests/minute

Enterprise: Unlimited

Rate limit headers: X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset