Claude Fable 5.1: Costs Drop 45%, Agent Benchmarks Double
kimmonismus · x · 2026-09-02
Anthropic released Claude Fable 5.1 with the following updates:
Pricing & Cost
- I/O pricing remains $10/$50 per million tokens. Cache reads drop 75% to $0.25.
- Estimated 25% cost reduction for typical workloads, up to 45% for heavy agentic tasks.
- Note: Subscription plans do not see the same price drop.
Performance Gains
- Scientific Agent: 52.6% (vs 24.7%), more than 2x improvement.
- AutomationBench: 31.4% (vs 17.1%), an 84% relative gain.
- CursorBench: 73.4% (vs 70.5%).
- Strong gains on long-running tasks, modest improvements elsewhere.
- More concise output; reports indicate half the token usage with double the speed.
Safety Policy
- 60% fewer cyber-safeguard interventions.
- Biology safeguards trigger 85% less often on benign queries.
- Vulnerability discovery is now allowed, but exploit generation and penetration testing remain restricted.
The release clearly targets enterprise users.
Related event: Anthropic Releases Fable 5.1 and Mythos 5.1 with 75% Cheaper Caching(174 posts)→
More from coding & agent
- Matt Pocock's pre-commit hook caps agent-written files at 5,000 tokens — mattpocockuk · 2026-10-18
- Author Laurence Moroney: write your ugly first draft yourself, then let an LLM edit it — lmoroney · 2026-10-18
- Open-weight TYGER speech decision model makes 80ms turn-taking calls for full-duplex voice agents — solyarisoftware · 2026-10-18
- Devs: Opus 5.5 on medium beats Opus on max — eyishazyer · 2026-10-18
- He Stopped Telling Claude Code to Stop Making Up Timestamps—and Made It Impossible with a PreToolUse Hook — Grimmoner · 2026-10-18
- Cursor Super + Claude Pro + OpenCode: can this stack survive heavy autonomous coding? — DarioAmodeiWasTaken · 2026-10-18