Fable 5.1 Early Data Shows an Agent-First Model, Not a Chat Model

Requesty ·

Key Info

Early usage data for Fable 5.1 (released Sep 1) shows it behaves more like an agentic coding model: median responses are much shorter, tool calls are up, and reasoning appears on about a third of requests. Coding agents dominate current usage, with Claude Code accounting for roughly 3 in 4 requests, though only 4% of teams that used Fable 5 in August have switched so far.

Highlights

  • Fable 5.1's median response dropped from 603 tokens to 270 tokens — fewer explanations, more action.
  • 78% of Fable 5.1 responses end in a tool call, versus 51% for Fable 5; reasoning tokens appear on 33% of requests, up from 27%.
  • Current usage is heavily driven by coding agents: Claude Code makes up about 75% of requests, while only 4% of prior Fable 5 teams have migrated.
Loading...