Anthropic's Fable 5.1 hits 90% on ARC-AGI-2 at 32% lower cost per task than Fable 5
rohanpaul_ai · x · 2026-09-03
ARC Prize published verified scores for Anthropic's Fable 5.1:
- ARC-AGI-2: 90.0% at $3.12/task, remarkably close to saturation
- ARC-AGI-1: 97.5% at $1.40/task
The average cost per task is 32% lower than Fable 5 across both benchmarks, driven by better token efficiency rather than reduced accuracy.
Related event: Anthropic's Fable 5.1 Scores 90% on ARC-AGI-2, Nearing Saturation(2 posts)→
More from Models
- CursorBench ranks coding agents: Fable 5.1 Max tops at 73.4%, Grok 4.6 best value — mattyp · 2026-09-03
- Google launches Agentic Video in Gemini, cutting video token usage by up to 88% — CSProfKGD · 2026-09-03
- Users say Gemini 3.8 Flash regressed for agentic tasks, calling 3.7 Flash calmer and faster — Scobleizer · 2026-09-03
- Scale CEO Praises Muse Spark 1.3's Interactive 3D Japanese Garden Scene as Huge Leap — alexandr_wang · 2026-09-03
- AA Predictions List Sparks Buzz With Mysterious 'DeepSeekV5-Preview: 58' Entry — teortaxesTex · 2026-09-03
- Gemini 3.8 Flash praised for quality but keeps hitting infinite loops in Cursor — brandon_galang · 2026-09-03