Skip the benchmarks. Run your own.
Quick eval
Pick models, a judge and a prompt, then start. Results appear right here.
Prompts
Search prompts...
All TypesCodingWritingReasoningQ&AInstruction followingLong contextRoleplay
- Implement binary search in PythonCodingWrite the function, then state its time complexity and edge cases.
- Turn the release notes into a launch blurbWritingKeep it under 80 words and preserve what actually changed.
- Three switches, three lampsReasoningYou may enter the room once. Work out which switch controls which lamp.
- Explain a vector database in plain languageQ&ASay what problem it solves, then give one real use.
Same prompt, answers side by side. Who starts first, and who finishes. The judge is never a model under test. Each score keeps the reason behind it. A public prompt, or one of your own. A real task, not a canned board.