ARC Prize Verified

Kimi K3

Moonshot AI·Jul 16, 2026·3 reasoning variants

As of July 31, 2026, Moonshot AI's Kimi K3 sets a new high score for open-weight models evaluated by ARC Prize on both ARC-AGI-1 and ARC-AGI-2. At max effort, it scores 94.5% on ARC-AGI-1 Semi-Private at $0.77 per task and 60.4% on ARC-AGI-2 Semi-Private at $1.59 per task. ARC-AGI-3 testing is still in progress.

ARC-AGI 2 leaderboard

Kimi K3

Verified scores

VariantARC-AGI-1ARC-AGI-2ARC-AGI-3
Max
94.5%
60.4%
High
86.7%
55.0%
Low
65.7%
12.4%

Tasks & environments

Pass/fail per reasoning level across each benchmark.

ARC-AGI-2 Public Eval

120 tasks
Task
Max
High
Low

ARC-AGI-1 Public Eval

400 tasks
Task
Max
High
Low