Download Data & Embed Widgets

<iframe src="https://benchlm.ai/embed/compare?a=claude-fable-5-1&b=gpt-6-astra&theme=light" width="100%" height="622" frameborder="0" title="BenchLM: Claude Fable 5.1 vs GPT-6 Astra"></iframe>
<p style="font-size:12px;text-align:right;margin-top:4px"><a href="https://benchlm.ai/compare/claude-fable-5-1-vs-gpt-6-astra" target="_blank" rel="noopener">Claude Fable 5.1 vs GPT-6 Astra benchmark comparison</a> by BenchLM.ai</p>

Live preview

The widget shows both public scores with their evidence status and rank, then the eight category rows with the same reading the compare page gives. Compare pages also offer CSV, JSON, and PNG exports directly.

Chart Widgets

Token Price Index

The frontier, mid-tier, and budget price index as a live chart.

<iframe src="https://benchlm.ai/embed/price-index" width="100%" height="360" frameborder="0" title="BenchLM Token Price Index"></iframe>
<p style="font-size:12px;text-align:right;margin-top:4px"><a href="https://benchlm.ai/token-price-index" target="_blank" rel="noopener">LLM token price index</a> by BenchLM.ai</p>

Context window chart

Documented context windows for the default model set.

<iframe src="https://benchlm.ai/embed/context-window" width="100%" height="460" frameborder="0" title="BenchLM context window chart"></iframe>
<p style="font-size:12px;text-align:right;margin-top:4px"><a href="https://benchlm.ai/tools/context-window" target="_blank" rel="noopener">LLM context window comparison</a> by BenchLM.ai</p>

API Documentation

GET /api/data/leaderboard

Returns ranked benchmark scores for all models. Supports category filtering and CSV export.

category — Optional. Filter by category: coding, agentic, reasoning, knowledge, math, multimodalGrounded, multilingual, instructionFollowing.

limit — Number of results (default: 50, max: 200).

format — Response format: json (default) or csv.

BenchAlign rows also carry interval90, categoryRanks, and categoryEvidence.

Example requests:

GET /api/data/leaderboard
GET /api/data/leaderboard?category=coding&limit=10
GET /api/data/leaderboard?format=csv

Response shape (JSON):

{
  "lastUpdated": "September 4, 2026",
  "models": [
    {
      "rank": 1,
      "model": "Claude Fable 5.1",
      "creator": "Anthropic",
      "sourceType": "Proprietary",
      "overallScore": 82.95,
      "categoryScores": {
        "agentic": 78.67,
        "coding": 84.23,
        "reasoning": 79.4,
        "multimodalGrounded": null,
        "knowledge": 86.39,
        "multilingual": null,
        "instructionFollowing": null,
        "math": null
      },
      "inputPrice": 10,
      "outputPrice": 50
    }
  ]
}

GET /api/data/pricing

Returns pricing information for all tracked LLM APIs.

limit — Number of results (default: 100, max: 200).

format — Response format: json (default) or csv.

Example requests:

GET /api/data/pricing
GET /api/data/pricing?format=csv&limit=50

Response shape (JSON):

{
  "lastUpdated": "April 3, 2026",
  "models": [
    {
      "model": "GPT-5",
      "creator": "OpenAI",
      "inputPrice": 2.50,
      "outputPrice": 10.00,
      "contextWindow": "128K",
      "sourceType": "Proprietary"
    }
  ]
}

GET /api/data/compare

Returns one comparison: both public scores with evidence status and rank, the eight category rows with their lane, basis, and reading, and every shared or single-model benchmark row with its source.

a, b — Two public model slugs (any order), or slug — a compare page slug such as claude-fable-5-1-vs-gpt-6-astra.

format — json (default) or csv.

GET /api/data/compare?a=claude-fable-5-1&b=gpt-6-astra
GET /api/data/compare?slug=claude-fable-5-1-vs-gpt-6-astra&format=csv

GET /api/data/benchmarks

Returns every public benchmark row for one model with its verification status and source, plus the public-lane category scores and ranks shown on the model page.

model — Required. A public model slug.

format — json (default) or csv.

GET /api/data/benchmarks?model=gpt-6-astra
GET /api/data/benchmarks?model=gpt-6-astra&format=csv

Data Freshness & Refresh Cadence

Both endpoints return a lastUpdated field in the JSON response. Use this to check data freshness.

Update frequency: Benchmark scores and pricing data are updated 1-3 times per week as new models launch, scores are verified, or providers change pricing.

Recommended refresh interval: If you cache API responses locally, re-fetch at least once per day. The API serves from a 1-hour CDN cache, so you will always get data less than 1 hour old.

Rate limiting: Free-tier API responses include a 3-second delay. Responses are cached at the CDN for 1 hour. For high-frequency or low-latency access, contact us.

Embed widgets: Embedded iframes fetch live data on each page view, so they stay current automatically with no action needed from you.

LLM Rankings

See the full benchmark leaderboard

LLM Pricing

Compare all LLM API prices

Price vs Performance

Find the best value AI models

Frequently Asked Questions

Can I use BenchLM data in my own analysis or product?

Yes. The JSON and CSV endpoints are free for personal and commercial use. We ask that you include attribution ("Data from BenchLM.ai") when publishing or displaying the data publicly. The embed code includes a pre-built attribution link — please keep it visible.

How do I customize the embed widget appearance?

Use the configurators above to choose a category, row count, and theme, or pick two models for a comparison widget. The generated iframe code reflects your choices and its height follows the row count. You can still adjust the iframe width and height attributes to fit your layout. The attribution link text adjusts automatically to match your selection.

How often is the data refreshed?

Benchmark scores and pricing data are updated whenever new models launch or prices change — typically 1-3 times per week. The API response includes a "lastUpdated" field so you can check freshness. CSV/JSON downloads always serve the latest data. If you cache API responses, we recommend refreshing at least once per day.

Are there rate limits on the API endpoints?

Yes. API responses include a 3-second intentional delay for free-tier access. Responses are also cached for 1 hour at the CDN layer. For faster or high-frequency access, contact us about priority API access.

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.

Browse BenchLM

Tools & calculators

Tools