Claude Fable 5.1 Keeps Its Price but Cuts Cache Costs

Anthropic released Claude Fable 5.1 on September 1, 2026, as the generally available successor to Fable 5. The change is less about a lower headline price than about reducing the cost of repeated context. LLM Stats reports that cache hits now cost $0.25 per million tokens, down from $1, while standard pricing remains $10 per million input tokens and $50 per million output tokens.

That makes the model particularly relevant to long-running agents, Claude Code workflows, and applications that repeatedly reuse prompts or tool context. Anthropic says typical workloads could be about 25% cheaper, with highly agentic usage potentially 45% cheaper, though those are indexed-cost estimates rather than independent measurements.

Capability changes

Anthropic’s self-reported launch table shows Fable 5.1 scoring 52.6 on Terminal-Bench-Science, compared with 24.7 for Fable 5, and 55.8 versus 42.0 on Terminal-Bench 4.0. AutomationBench also rose from 17.1 to 31.4. Improvements were more modest on CursorBench, where the model reached 73.4 versus 70.5, and on HLE with tools, at 65.0 versus 63.8.

The benchmark results include production safeguards and have not been verified by LLM Stats. Terminal-Bench-Science carries an error range of roughly four points, so the result should be treated as a performance band rather than an exact ranking.

Fable 5.1 supports a 1,048,576-token input context and 128,000-token maximum output. Cache writes cost $12.50 for five-minute retention or $20 for one hour. Batch pricing is half the standard rate. Cybersecurity and research-biology tasks may still fall back to more expensive Opus models, making workload routing important when estimating costs.

Source: LLM Stats