Skip to content

Sources & credits

Every source this site uses, what we use it for, its licence and how we credit it.

Data sources

  • BenchLM ↗
    Retrieved Oct 20, 2026

    Used for: Overall and category scores with 90% intervals, evidence labels, per-benchmark results with citations, list prices.

    Licence: CC BY-NC 4.0

    How we credit it: "Data from BenchLM.ai" with a link, on this page, linked from the menu on every page. Non-commercial use only.

  • OpenRouter ↗
    Retrieved Oct 20, 2026

    Used for: Hosted prices, context lengths, input/output types, and models not yet benchmarked.

    Licence: Public API, used with attribution

    How we credit it: Named and linked on this page and as the price source on each model's pricing table.

  • BenchLM change feed ↗
    Retrieved Oct 20, 2026

    Used for:

    Licence: CC BY-NC 4.0

    How we credit it: Data from BenchLM.ai

  • LMArena ↗
    Retrieved Oct 20, 2026

    Used for: Human-preference arena ratings: text, webdev, vision, search, documents, text-to-image, image editing, text-to-video, image-to-video.

    Licence: CC BY 4.0

    How we credit it: "Ratings from LMArena" with a link, on this page, linked from the menu on every page.

  • Hugging Face ↗
    Retrieved Oct 20, 2026

    Used for: Parameter counts, licences, downloads and release dates of open-weights models.

    Licence: Public API, used with attribution

    How we credit it: Each model page links to its Hugging Face repository.

Official lab pricing pages

Prices marked ✓ Verified were checked by hand on these pages; every refresh compares them with the other sources.

Anthropicplatform.claude.com/docs/en/about-claude/pricing16 models · checked Oct 17, 2026
OpenAIdevelopers.openai.com/api/docs/pricing
Standard tier, prompts under 272K tokens.
21 models · checked Oct 17, 2026
Googleai.google.dev/gemini-api/docs/pricing
Paid tier, prompts up to 200K tokens. Several prices change on Dec 1, 2026 / Jan 1, 2027.
12 models · checked Oct 17, 2026
xAIdocs.x.ai/docs/models
Prompts under 200K tokens.
5 models · checked Oct 17, 2026
DeepSeekapi-docs.deepseek.com/quick_start/pricing
Off-peak price; peak hours (01:00-04:00 and 06:00-10:00 UTC, weekdays) cost 2x.
2 models · checked Oct 17, 2026

Our own measurements

Speed (time to first token, output speed, end-to-end time) will be measured by ModelBench through OpenRouter; no measurements are published yet. We never copy speed figures from other sites.

Logos

Company logos come from LobeHub Icons (MIT licence). The logos themselves are trademarks of their owners and are shown only to identify each model's maker; no endorsement is implied.

Using our data

Compiled data: CC BY-NC 4.0 (includes data from BenchLM.ai). Speed measurements: ours. Credit "ModelBench", "Data from BenchLM.ai" and "Ratings from LMArena" with links. Last compiled Oct 20, 2026.