On Tuesday, September 1, 2026, Anthropic released Claude Fable 5.1 and Claude Mythos 5.1, the latest versions of its most advanced large language models. The company positions the two models—which share the same underlying architecture—as the “world’s most advanced models for coding and knowledge work.” The distinction lies in their safeguards: Fable 5.1 is the generally available version with Anthropic’s standard production safeguards, while Mythos 5.1 is a restricted version available only through trusted access programs for vetted partners in cybersecurity and life sciences.
A key feature of the release is a significant cost reduction. While the base pricing for Fable 5.1 remains unchanged at $10 per million input tokens and $50 per million output tokens, Anthropic has reduced the price of cache reads by 75%. This leads to an estimated 25% lower cost for typical workloads and up to 45% savings for highly agentic tasks.
Anthropic reports substantial performance improvements for the new models across several benchmarks. On Terminal-Bench-Science 0.1, which evaluates agentic scientific research, Fable 5.1 scored 52.6%, more than doubling Fable 5’s score of 24.7%. For coding on Terminal-Bench 4.0, Fable 5.1 scored 55.8%, outperforming Opus 5 (52.3%) and OpenAI’s GPT-5.6 Sol (37.3%). When operating under its more permissive safeguards, Mythos 5.1 reached 60.9% on the same benchmark. Beyond these scores, Anthropic highlights that Fable 5.1 demonstrates “better judgment on ambiguous tasks and fewer confident wrong answers.” The model’s knowledge cutoff has also been updated to June 2026.
The release also introduces refined safety and security features. Anthropic has improved the model’s safeguards to reduce false positives; for instance, cybersecurity interventions in Claude Code sessions are now approximately 60% fewer. Fable 5.1 is now permitted to help identify software vulnerabilities, though it remains unable to build exploits for them. To protect against model distillation, new API accounts can no longer edit prior conversation history to preserve Claude’s chain-of-thought, a change aimed at preventing rivals from training cheaper models on Anthropic’s outputs. Additionally, Fable 5.1 includes invisible watermarking for generated text, with a detection API in private preview.
The models are available on Anthropic’s platform, Amazon Bedrock, Google Cloud, and Microsoft Foundry. Fable 5.1 is generally accessible, while Mythos 5.1 is limited to U.S.-based partners in cybersecurity and life sciences through a trusted access program. A new Life Sciences Verification Program will soon allow scientists to apply for access to Mythos 5.1’s advanced biology capabilities. Anthropic has also announced that Fable 5.1 will not be retired before September 1, 2027.
Email your news TIPS to Editor@Kahawatungu.com — this is our only official communication channel
