Claude Fable 5.1's five effort levels span 11x in token usage, 58-66 on AA Index
ArtificialAnlys · x · 2026-09-02
Artificial Analysis measured Claude Fable 5.1's five effort settings spanning 11x in output tokens — from 13.1M at low effort to 143.7M at max — with Intelligence Index scores from 58 to 66.
All effort levels sit on the Intelligence vs Output Tokens frontier, but Fable 5.1's floor is higher than GPT-5.6's: GPT-5.6 Sol (medium) uses 12M tokens, slightly fewer than Fable 5.1's lowest. On cost, Fable 5.1 (max) runs $3.76 per task, 20% above Fable 5 ($3.14) and 1.6x Opus 5 ($2.34), driven by 1.7x the output tokens. Anthropic's cache read price cut from $1 to $0.25/1M tokens saves $1.40 per task; without it, cost would be $5.16. At xhigh, it scores 65 at $2.72 per task, $1.04 less than max.
More from Models
- MSL Releases Muse Voice: SOTA Real-Time Transcription Model — bowenc0221 · 2026-09-02
- Fable 5.1 strong at biology but frequent refusals limit utility — kenbwork · 2026-09-02
- Anthropic allows Fable 5.1 for vuln scanning, but prompt triggers downgrade — AccBalanced · 2026-09-02
- ChatGPT Remains Sycophantic While Logged In; Free and Paid Versions Show Significant Differences — lilyraynyc · 2026-09-02
- Fable 5.1 system prompt leaked by jailbreaker Pliny within an hour of release — Polymarket · 2026-09-02
- Reddit Predicts Open-Weight Models Won't Match Fable 5.1 Until Late 2025 — power97992 · 2026-09-02