benchr
benchr compiles reference information on AI language models — pricing, context windows, benchmark scores, deprecation status — sourced from official provider documentation. Coverage includes Claude, GPT, Gemini, Llama, Mistral, DeepSeek, Qwen, Phi, and other frontier and open-weight models.
The publication is structured as a working reference, not as a magazine of original research. Articles describe each model's documented capabilities, published pricing, and known limits, with analysis on which model fits which workload based on the public record. Where original evaluation is presented, the test setup is described in the article itself.
What you'll find
Pricing tables by model and provider, sourced from Anthropic's pricing page, OpenAI's API pricing, and Google's Gemini API models page among others. Benchmark scores referenced to the original benchmark maintainers — SWE-bench Verified, LMSYS Arena, ARC-AGI. Context-window comparisons. Notes on deprecations, version histories, and which model has replaced which. Recommendations on which model fits which use case, sourced from the published capabilities of each.
What you won't find
Synthetic precision claims. Invented test costs. Sponsored content. Affiliate links to providers. Empty doorway pages built to catch a search query without answering it.
benchr does publish use-case guides that target real buyer questions, like "best AI for X" comparisons. Those exist because people genuinely search them, but each one carries its own reporting: a specific pick, real pricing, and a different recommendation per use case, not a template with the words swapped. The line we hold is utility. A page has to earn its place by answering the question better than a generic listicle would.
Use of AI
benchr is independent — no AI provider sponsors it, and it takes no affiliate fees from AI providers. AI tools are used to draft and update articles; every factual claim (pricing, benchmarks, dates) is checked against the provider's primary source before publishing.
Article datelines distinguish two things: "Updated" means the content itself changed, logged in that article's changelog, while "Reviewed" means the facts were re-checked against primary sources on that date without the content changing. A site-wide verification pass ran on May 30, 2026.
Corrections
Errors are corrected on the original article with a note in the article's changelog. Material corrections also appear on the corrections page.
Contact
For corrections or source disputes: corrections@benchr.org.