Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1, arriving just three months after the original Fable 5 line launched in June 2026. Both models share the same underlying architecture but are deployed with distinct safety protocols. Claude Fable 5.1 is now generally available as claude-fable-5-1, while Claude Mythos 5.1 remains restricted to vetted organizations.
Both models feature a 1 million token context window and support up to 128,000 output tokens, with adaptive thinking always enabled. The standout performance metric is 52.6% on Terminal-Bench-Science 0.1, a significant jump from 24.7% for Fable 5 and 29.0% for Opus 5. On the commercial side, cache read pricing has dropped by 75%, from $1.00 to $0.25 per million tokens—a move that Anthropic estimates will lower typical workload costs by roughly 25% and agentic workloads by up to 45%. Base input and output pricing remains unchanged at $10 and $50 per million tokens, respectively.
Is It Deployable?
Yes, Claude Fable 5.1 is generally available as claude-fable-5-1 through the Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry. In contrast, Claude Mythos 5.1 is not publicly accessible; it is limited to vetted US organizations participating in Project Glasswing.
Benchmark Performance
On Terminal-Bench-Science 0.1, an agentic scientific research benchmark, Fable 5.1 achieves 52.6%, outperforming Opus 5 (29.0%), Fable 5 (24.7%), and GPT-5.6 Sol (22.4%). Anthropic notes a standard error of 3.5 to 4.5 points per model, advising that the ranking order is reliable, though the exact margins should be interpreted cautiously.
For broader context, on Terminal-Bench 4.0, the model continues to show strong capabilities, though detailed results were not fully disclosed in this release. These improvements underscore Anthropic's focus on advancing agentic AI performance while also reducing operational costs, making advanced AI more accessible for enterprise and research applications.
via MarkTechPost