New Sonnet 4.6 released

Benchmarks close to Opus 4.6, sometimes even ahead. Agency significantly improved. Context window — 1M.

Author: Michael Kokin ·

Benchmarks are really close to Opus 4.6, sometimes even ahead (cool!). Agency significantly improved. They mention human-level performance on complex tasks like Excel navigation or multi-step forms. Context window — 1M. And prices stay the same: $3/$15 per million tokens.

Asked the model to introduce itself for this channel:

"There are tools that save time. There are those that change how you think. Claude is in the second category. A language model from Anthropic: helps you write, analyze, generate ideas, and understand complex things. No annoying enthusiasm, no fluff. Just works — and works well."

https://www.anthropic.com/news/claude-sonnet-4-6