Anthropic cuts Claude Fable 5.1 cache read costs by 75%
Anthropic announced the release of Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026. While the two models share the same underlying architecture, Mythos 5.1 is restricted to trusted organizations in high-risk fields like cybersecurity and life sciences, whereas Fable 5.1 is available to the general public. The most significant change is a 75% reduction in the cost of "cache reading,"…
Key points
- Anthropic reduced cache read rates for Claude Fable 5.1 by 75%.
- Long-duration coding task costs are expected to drop by up to 45%.
- Fable 5.1 outperforms Opus 5 and GPT-5.6 Sol in agentic benchmarks.
According to Anthropic, the new models show improved performance in long-running tasks and autonomous tool usage. Benchmarks indicate that Fable 5.1 outperforms the previous Fable 5, Anthropic's own Opus 5, and OpenAI's GPT-5.6 Sol in automated software development and agentic tasks. The company also claims the model has a stronger tendency to address root causes of issues rather than applying superficial patches, which may reduce errors in complex, autonomous workflows.
This pricing strategy aligns with broader industry trends where competition drives down costs for high-volume usage. For users relying on Claude for bulk coding or long-context interactions, the update offers immediate cost savings without requiring any changes to their setup, as the models update automatically.
Model pages: Claude Fable 5.1 → · GPT-6 Astra →
The story so far
4 episodes →- Anthropic cuts Claude Fable 5.1 cache read costs by 75%this story
Claude actually lowered its prices. What changed in Fable 5.1?
note.com · 20 September 2026
Loading the full article…
This text was published by note.com and written by sho AI最新情報まとめ. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- JS Denain of Epoch AI says Chinese models lag US labs by about six to eight months · 1 src
- Aws preview TBC rat‑Brain video model for select customers · 1 src
- Xiaomi releases MiMo-V2.6-Pro-RL, a 1.02T/42B open omnimodal model trained via scaled RL · 5 src
- AI Safety Concerns Erupt Amid Recursive Self-Improvement Fears · 3 src
- Tencent launches Hy Image 3.5 preview, says it matches ByteDance's Seedream 5.0 Pro · 1 src
Comments
via GitHub Discussions