The craziest thing about Fable 5.1 for me personally
Summary
The author used Claude and ccusage to calculate that 78% of costs for Fable 5.1 come from cache reads, predicting a 57% cost reduction if cache reads become cheaper.
Similar Articles
For agent loops the cache read discount is the whole story on Claude Fable 5.1
The article compares Claude Fable 5.1 and 5, showing a 7.5% cost reduction for long agent loops due to a 75% discount on cache reads, which dominates billing in extended sessions.
So much for Fable 5.1 being cheaper. Its cost per task is higher than Fable 5 at $3.69
Fable 5.1 has a higher cost per task than Fable 5, contrary to initial expectations of it being cheaper.
Making Fable Cheaper Than Opus (12 minute read)
Cognition replaced the Opus model with Fable in Devin's Fusion architecture, achieving higher performance at lower cost despite Fable's higher per-token price, through better delegation and reduced lead model turns.
@natolambert: The crazy jump in perf for Claude 5 Fable is vindication for people who say Opus 4.5 and were like "yeah I should (most…
Nathan Lambert highlights the significant performance improvement of Claude 5 Fable, suggesting it validates the shift away from manual coding.
@PrajwalTomar_: WAIT. Anthropic just proved you don't need Fable 5 for everything. With their own benchmarks. They tested "Fable 5 orch…
Anthropic benchmark shows that using a larger model (Fable) as orchestrator with cheaper models (Sonnet) as workers achieves 96% of full Fable performance at 46% cost, available now in Claude Code.