Anthropic Releases Claude Fable 5.1, Doubling Performance Amidst Growing Frustration Over Usage Limits

In the "Terminal-Bench-Science 0.1" benchmark targeting scientific research tasks, Claude Fable 5.1 recorded 52.6%, a nearly twofold increase from the 24.7% recorded by the previous model, Fable 5.

In "Terminal-Bench 4.0", which measures software development tasks, performance also rose from 42.0% to 55.8%.

Regarding pricing, the company has reduced the cost of cache reads (a mechanism to reuse previously loaded conversation content) by 75%. Anthropic explains that this can lead to cost savings of up to 45% for agentic use cases.

On the other hand, voices of dissatisfaction regarding usage limits are rising on the social media platform X. Users subscribed to Anthropic's premium plans have reported, "After about 45 minutes of coding, I reached the 5-hour session limit and consumed 38% of my weekly quota." Additionally, multiple posts have been confirmed stating that limits are reached within minutes, even under low-load settings.