Claude Fable 5.1 Launch, Cache Costs Cut 75%

The launch comes as other models like Grok 4.7 and Gemini Flash 3.8 approach, but Fable 5.1's reception highlights a split between raw capability and practical use. Benchmarks show large gains, such as doubling Terminal-Bench-Science scores, but users report rate limits and ergonomic issues. The price cut on cache reads signals a focus on long-running agentic tasks, yet the higher output token usage tempers the benefit. Artificial Analysis notes Fable 5.1 may tie with Opus 5 on some agentic knowledge work, so the win is not absolute. Community questions about whether Fable and Mythos are distinct models add uncertainty to benchmark interpretation.