Pricing, benchmarks, and use-case fit for Claude, GPT, Gemini, Llama, Mistral, and the open-weight tier — sourced from official provider documentation.
Featured
A head-to-head comparison covering code review, landing-page copy, citation accuracy, recipe writing, and email drafting. Claude leads on five of the seven; the two GPT-5 wins are not small.
OpenAI's new series ships behind a government gate — about 20 partners only. Announced at $5/$30, $2.50/$15, and $1/$6, with context windows and API IDs still unpublished.
The Mythos-class model the public could finally use — until a June 12 U.S. export-control order suspended it. $10/$50, falls back to Opus 4.8.
Same price as 4.7, a small leaderboard bump, and an honesty gain that catches its own bugs.
OpenAI's gains land in agentic coding and computer use, at roughly double GPT-5's API price.
Both charge $5 per million input. Where each one wins after that, and which belongs in your stack.
Anthropic's restricted frontier model, gated under Project Glasswing. What it is and why it's locked away.
Strong on architecture decisions and long-document analysis. Vision is its weaker side.
The visual-design winner. Still strong on math and structured output.
Excels at vision + reasoning. Average elsewhere. Deprecated by Google in March 2026 in favor of 3.1 Pro.
The cheapest model for chat, coding, RAG, agents, classification, and summarization.
Where Llama 4, Mistral Large 2, DeepSeek-V3.1, and Qwen 3 stand against the closed labs.
MMLU is saturated above 90%. The benchmarks worth tracking now.
Claude Opus 4.7, GPT-5, and Gemini 3.1 Pro Preview: the reviews, the head-to-head, and a single buying decision.
Llama 4, Mistral, DeepSeek, Qwen, the small-model tier, and what it takes to self-host.
What AI costs by model and workload, and where teams overspend.
An interactive reference covering pricing, benchmarks, context windows, and capability ratings for all 21 models in the index, across the frontier, mid, and open-weight tiers.
All 26 models ranked by benchr Rating →
benchr is a verified reference, not a blog. Every price, benchmark, context window, and date is checked against the provider's own official documentation — sourced and dated — before it goes on the page.
benchr