Building custom evals shouldn't be hard or take forever
You can use our plugin + MCP to create a full eval set, then run it across treatments, models, or agent harnesses you care about we tried with firecrawl. we wanted to understand how well their MCP behaves and the lift it
Claude Opus 5 is crazy for web design.
Prompt ↓
Fable 5.1 is now available in Cursor - it takes the top spot on CursorBench.
Updated scores: 1. Fable 5.1 Max: 73.4% 2. Fable 5.1 Extra High: 72.8% 3. Grok 4.6 Extra High: 70.8% 4. Fable 5 Max: 70.5% 5. Opus 5 Max: 70.0% 6. Grok 4.6 High: 69.9% 7. Fable 5.1 High: 69.4% 8. Opus 5
Fable 5.1 UI gave me goosebumps, first of all models I've tried so far.
Opus 5 is dethroned (which in my opinion was better than Fable 5 at this). This is truly single shot. None of the specific animations it used are in the prompt either. See for yourself, link below.
You have reached the end of the archive
All of Claude Opus 5