Results

Results

Model performance across ARC-AGI benchmarks

DateProviderModelARC-AGI-1ARC-AGI-2ARC-AGI-3
Aug 13, 2026DeepSeekDeepSeek V4 Pro 081390.5%61.3%
Aug 13, 2026GoogleGemini 3.7 Flash95.5%84.6%
Aug 11, 2026xAIGrok 4.687.5%67.1%2.11%
Jul 31, 2026DeepSeekDeepSeek V4 Flash 073189.0%61.4%
Jul 30, 2026OpenAIGPT-5.6 Luna 2026-07-3090.7%59.6%
Jul 30, 2026Thinking MachinesInkling Small84.0%40.1%
Jul 24, 2026AnthropicClaude Opus 597.5%90.4%30.16%
Jul 21, 2026GoogleGemini 3.5 Flash-Lite53.5%10.3%
Jul 21, 2026GoogleGemini 3.6 Flash91.2%60.4%
Jul 16, 2026xAIGrok 4.587.2%52.6%0.32%
Jul 16, 2026Moonshot AIKimi K394.5%60.4%
Jul 15, 2026Thinking MachinesInkling79.5%36.5%
Jul 9, 2026OpenAIGPT-5.6(Sol · Terra · Luna)97.5%92.5%7.78%
Jun 9, 2026AnthropicClaude Fable 598.5%89.2%