benchr
Choose a model for your work
Answer four questions about the task, price sensitivity, priority, and deployment. The tool returns three candidates from the index and shows why each one scored well.
Other tools
Frequently asked questions
What's the best AI model for coding?
There is no universal coding winner. The recommender calculates an editorial fit score from your selected task, budget, and deployment preferences; that score is not a measured benchmark. Treat Claude Opus 4.8, DeepSeek V4-Pro, and relevant open-weight models as candidates, then test them on the same repository tasks before choosing.
What's the best AI model for writing?
There is no universal writing winner. The tool's score is an editorial fit calculation, not a blind writing test. Shortlist the models that meet your budget and deployment constraints, then compare them on the same brief with blinded reviewers for voice, factual discipline, required edits, and long-document consistency.
Which model should I self-host?
The right self-hosted model depends on hardware, license, language, context, and workload. Use the privacy filter to produce an editorial shortlist, verify every model's current license and memory requirements, and benchmark the finalists on your own hardware and tasks.
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.