Anthropic's Fable 5 Sets New High Score on ARC-AGI Benchmarks
mhmazur · x · 2026-08-06
ARC Prize officially verified the performance of Anthropic's Fable 5 on the ARC-AGI benchmarks. The model achieved a high score of 89.2% on ARC-AGI-2 ($5.45/task) and 98.5% on ARC-AGI-1 ($1.02/task), making it the highest-scoring model ever evaluated on ARC-AGI-1 to date.
More from Models
- ARC-AGI 3 benchmark fallen; Ilya's SSI model crushes it harness-free — iruletheworldmo · 2026-08-06
- Muse Spark 1.2 Scores 54 on Intelligence Index with Competitive Pricing — ArtificialAnlys · 2026-08-06
- Meta's Muse Spark 1.2 Boosts Agentic Capabilities, Ties for 3rd Among US Labs — ArtificialAnlys · 2026-08-06
- Next-gen LLMs leaked: Qwen3.8 Max, DeepSeek V4 Pro pricing revealed — togethercompute · 2026-08-06
- Meta API Pricing Revealed: Contributor Tier Drops Input Costs by Up to 98% — AIatMeta · 2026-08-06
- Hands-on: Muse Code Shines in 3D Visual Agentic Coding Task — cedric_chee · 2026-08-06