Claude Fable 5.2 on FarmerBench
80.16, second to Gemini 4 Argon (84.01), ahead of Opus 5.5 (75.54) and Fable 5.1 (69.17). 6 game-hours to automate a farm economy by writing code. Fable 5.2 hit 32 drones on every seed.
Yes. I see where you are going and the distinction is real: coding is a ladder and voice is a taste.
1. **Coding has one direction of better.** A test passes or fails. The user sees the result and never the words inside the code. So a newer model that wins every coding test can
Huge jump from Claude 🤯!!
Launching Fable 5.2 and Mythos 5.2 with performance superior to Opus 5.5 in most tests.. Most notable new features: 1- A significant improvement in programming, logical thinking, and understanding images. 2- Up to 30% faster generation. 3- Strong software performance at approximately half the cost. 4- Advanced control capabilities…
This is an actual f*cking goldmine
Anthropic just shared one of the best tricks for Claude Code. The idea: run Sonnet, Haiku 5.5 and Opus 5.5 at once, each one doing only the job it's best at. Better results and way fewer tokens, because every task lands on the right model.
You have reached the end of the archive
All of Claude Opus 5