LMArena
Crowd-voted head-to-head model rankings.
AI Lists βΊ AI Models & Benchmarks
Where to compare models: leaderboards, the benchmarks behind them, and each model family's official page.
Rankings change weekly, so this list points to the live leaderboards and the official model pages rather than copying scores that go stale.
Crowd-voted head-to-head model rankings.
Independent comparisons of model quality, speed and price.
Benchmark with fresh questions to limit contamination.
How well models edit code in Aider's polyglot benchmark.
Stanford's holistic evaluation of language models.
Massive Text Embedding Benchmark rankings.
57-subject multiple-choice knowledge test.
Graduate-level, 'Google-proof' science questions.
Very hard expert-written questions across many fields.
Abstract reasoning puzzles easy for people, hard for AI.
Unpublished research-level maths problems.
Competition mathematics problems with step-by-step solutions.
Grade-school maths word problems.
College-level multimodal understanding and reasoning.
Short factual questions to measure hallucination.
Can a model resolve real GitHub issues?
Classic Python function-writing benchmark.
Agents completing real tasks in a terminal.
Tool-using agents in customer-service style tasks.
Agents operating real desktop computers.
Agents completing tasks on realistic websites.
General AI assistant tasks needing tools and browsing.
Hard-to-find facts that require persistent web browsing.
Current OpenAI model list and capabilities.
Current Claude model list.
Current Gemini API models.
Google's open-weight model family.
Meta's open-weight model family.
Mistral's open and commercial models.
Alibaba's open model family on GitHub.
DeepSeek's open model releases on GitHub.
Data on AI compute, models and trends.
Measures how long a task AI agents can finish on their own.
For AI agents and tools: JSON Β· Markdown Β· llms.txt Β· RSS. Page built Nov 20, 2026, 12:17 AM PT.