Anthropic Releases Claude Fable 5.1 with New Performance Benchmarks
Verified
Anthropic has released Claude Fable 5.1, which they say sets a new standard for coding, knowledge work, and long-running problem-solving tasks. While the company publishes system prompts for models like Haiku 4.5, the prompts for Claude Cowork and Code remain unavailable to the public. The new model achieved a 52.6% score on Terminal-Bench-Science 0.1, a benchmark whose announcement coincided with the release on August 27. A system card for the update is accessible online, and a Hacker News discussion item regarding the release received 235 points.