Which AI model should I use?
Answer four quick questions. We weigh 327 models on quality, price and features, and suggest the three that fit you best.
Our suggestions
Small models are often noticeably weaker in other languages than in English, so for French, Spanish and Arabic we favor stronger models. Text in these languages also uses more tokens, which is included in the cost estimate.
Quick answers
- Which AI model is best for coding?
- Claude Opus 4.7 (Anthropic) has the highest score on SWE-bench Verified (83.5%), a test built from real bugs in real software projects. See details →
- What is the best AI model on a small budget?
- GPT-5.6 Luna (OpenAI) has the best overall score among models costing at most $2 per million output tokens: it costs $0.20 / $1.20 per million tokens. See details →
- Which AI model is best for long documents?
- GPT-6 Astra (OpenAI) combines a 1,050,000-token context window, enough for about 1,969 pages at once, with a strong overall score. See details →
- What is the best open-weights AI model?
- Kimi K3 (Moonshot AI) has the best overall score among models whose weights are published, so you can run it on your own servers. See details →
How we choose
Suggestions combine independent quality scores (Epoch AI), current API prices and each model’s features. The monthly estimate assumes typical usage for the task you picked. It is only a guide: test two or three models on your own examples before deciding.