Claude Fable 5.1 and Claude Mythos 5.1

Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1 as its new flagship models designed specifically for coding, knowledge work, and autonomous multi-step problem solving. Both models feature a one-million-token context window with text and image inputs. Anthropic kept standard list pricing the same for inputs, outputs, and cache writes, but cut the cache-read price by 75 percent to twenty-five cents per million tokens. Third-party testing places Fable 5.1 at the top of the Artificial Analysis Intelligence Index and shows significant performance gains across various coding and reasoning benchmarks, including Terminal-Bench-Science and Humanity’s Last Exam.

Community reaction is heavily split between technical praise and practical friction. Developers and builders praise the model for superior planning, faster speed, better end-to-end agent workflows, and a more natural conversational tone with fewer AI stylistic tells. However, users report severe rate limits, subscription frustrations, and false-positive safety triggers that block legitimate technical tasks. Technical analysts also highlighted that evaluation runs included server-side fallback routing to older Opus models, and community speculation suggests Fable and Mythos 5.1 may share identical underlying weights separated only by safety classification thresholds.

This release matters because it shifts the focus from raw benchmark scores to practical, long-running agent economics and enterprise readiness. The substantial cache-read price cut makes multi-step agentic workflows much more affordable by reducing the cost of re-reading large codebases and scratchpads. Additionally, the introduction of zero-data-retention support and Enterprise Frontier Safeguards addresses critical corporate requirements for observability and data privacy, helping position the models for widespread commercial deployment despite ongoing user complaints regarding consumer access and subscription limits.