benchr
Find your model
Answer four questions about your task, budget, priority, and privacy requirements. benchr returns your top three picks with a one-line reason for each — sourced from the ranked index, no sponsorship.
Other tools
Frequently asked questions
What's the best AI model for coding?
Claude Opus 4.8 leads coding at 9.7/10 and is the top-ranked coding model in benchr's index. For budget-conscious coding, DeepSeek V4-Pro delivers near-Opus benchmark scores at under $0.44/1M input. For free self-hosted coding, Llama 4 Maverick is the pick. Use the recommender above for your specific budget and requirements.
What's the best AI model for writing?
Claude Sonnet 4.6 scores 8.9/10 for writing. Highest in the mid-price tier. GPT-5.5 handles long-form drafting well, especially documents over 50,000 words. Mistral Medium 3.5 is the top open-weight writing choice.
Which model should I self-host?
Llama 4 Maverick is the strongest open-weight option. DeepSeek V4-Pro (MIT) and Kimi K2.6 are good alternatives. Pick the privacy option in the recommender to filter only open-weight models.
Does this recommender use live routing?
No. It's a static recommendation from models.json based on your answers — no API call to any provider. The logic is in assets/js/recommender.js, which you can read and verify.