DeepSeek V4 Flash 0731 Scores 89% on ARC-AGI-1, 61.4% on ARC-AGI-2
DeepSeek published V4 Flash 0731, a model with three reasoning variants (Max, High, Low), reporting verified scores on ARC-AGI benchmarks. At max effort, it achieves 89.0% on ARC-AGI-1 Semi-Private at $0.02 per task and 61.4% on ARC-AGI-2 Semi-Private at $0.04 per task. The release includes a paper, model link, and per-task pass/fail results for the ARC-AGI-2 Public Eval.
Coverage timeline
Hacker Newstosh
# DeepSeek V4 Flash 0731 DeepSeek·Jul 31, 2026·3 reasoning variants Paper ↗Model ↗ At max effort, DeepSeek V4 Flash 0731 scores 89.0% on ARC-AGI-1 Semi-Private at $0.02 per task and 61.4% on ARC-AGI-2 Semi-Private at $0.04 per task. ## ARC-AGI 2 leaderboard DeepSeek V4 Flash 0731 ARC-AGI-1 ARC-AGI-2 ## Verified scores | Variant | ARC-AGI-1 | ARC-AGI-2 | ARC-AGI-3 | | --- | --- | --- | --- | | Max | 89.0% | 61.4% | — | | High | 87.0% | 56.0% | — | | Low | 84.0% | 46.0% | — | ## Tasks & environments Pass/fail per reasoning level across each benchmark. ### ARC-AGI-2 Public Eval 120 tasks | Task | Max ✓✗ | High ✓✗ | Low ✓✗ | | --- | --- | --- | --- | | 0934a4d8 | ✗ | ✓ | ✗ | | 135a2760 | ✓ | ✓ | ✓ | | 136b0064 | ✓ | ✓ | ✓ | | 13e47133 | ✗ | ✗ | ✗ | | 142ca369 | ✗ | ✓ | ✗ | | 16b78196 | ✗ | ✗ | ✗ | | 16de56c4 | ✓ | ✓ | ✓ | | 1818057f | ✓ | ✓ | ✓ | | 195c6913 | ✓ | ✗ | ✗ | | 1ae2feb7 | ✓ | ✓ | ✓ | | 20270e3b | ✓ | ✓ | ✗ | | 20a9e565 | ✗ | ✗ | ✗ | | 21897d95 | ✗ | ✗ | ✗ | | 221dfab4 | ✓ | ✓ |