Fable 5.1 cuts cache reads 75%, making agentic workloads ~45% cheaper overall

rohanpaul_ai · x · 2026-09-02

Rohan Paul relays a16z interview comments from Gavin Baker: frontier labs always hold back more advanced checkpoints — with Fable 5.1 out, OpenAI's Astra could drop anytime.

His earlier analysis of Fable 5.1 pricing: cache reads cost 75% less than Fable 5. Agents constantly revisit the same context, plans, tool outputs and working state, so a 75% cut on one component translates to roughly 45% off the entire workload. The core enterprise economics question is now "how long can an agent keep working before the job stops making financial sense" — so this cost curve must keep falling.

Related event: a16z Partner Predicts OpenAI's Astra Is Coming Soon(2 posts)→

Original post →

More from Models

Models channel →