UnifyBench

RankingARC-AGI-2

ARC-AGI-2

Data updated 22 Oct 2026

Bucket
Hard reasoning
Unit
percent
Direction
Higher is better
Version
2
Display harness
ARC Prize verified
Board
https://arcprize.org/leaderboard

Compare published benchmark results with category weights →

Models

Results are ordered by score in the display harness, followed by models with no result. Other harnesses are listed separately below. Missing results remain unknown.

1–25 of 471 entries

ModelScoreEvidenceSource-recorded date
GPT-6 AstraOpenAI95%official board2026-09-02
GPT-6.1 SolOpenAI94.2%official board2026-09-29
Claude Opus 5.5Anthropic93.3%official board2026-09-22
GPT-5.6 SolOpenAI92.5%official board2026-07-09
Gemini 4 ArgonGoogle91.3%official board2026-10-13
Claude Opus 5Anthropic90.4%official board2026-07-24
Claude Fable 5.1Anthropic90%official board2026-09-01
GPT-6 SolOpenAI89.6%official board2026-09-22
Claude Fable 5Anthropic89.2%official board2026-06-09
Gemini 3.8 FlashGoogle89.2%official board2026-09-02
GPT-5.5OpenAI85%official board2026-04-22
Gemini 3.7 FlashGoogle84.6%official board2026-08-13
GPT-5.5 ProOpenAI84.6%official board2026-04-23
Claude Sonnet 5.5Anthropic83.9%official board2026-09-28
GPT-5.6 TerraOpenAI83.9%official board2026-07-09
GPT-5.4 ProOpenAI83.3%official board2026-03-04
Gemini 3.1 ProGoogle77.1%official board2026-02-19
GPT-5.4OpenAI74%official board2026-03-04
DeepSeek V4.1 FlashDeepSeek72.9%official board2026-09-10
Claude Opus 4.8Anthropic72.1%official board2026-06-01
Gemini 3.5 FlashGoogle72.1%official board2026-05-19
Grok 4.6xAI67.1%official board2026-08-11
GLM-5.3-FlashZ.ai65.8%official board2026-08-26
Grok 4.7xAI61.4%official board2026-09-21
Claude Haiku 5.5Anthropic61.3%official board2026-10-07