Kimi K2 Thinking

Kimi K2 Thinking from Moonshot AI achieves 99.1% on AIME and 84.5% on GPQA, placing it among the top mathematical reasoning models in mid-2026. At $0.60/M, it delivers the best quality-per-dollar ratio in its capability class, undercutting comparable models by a significant margin. The ‘Thinking’ designation indicates extended chain-of-thought reasoning is applied by default, making it especially powerful for multi-step logical and quantitative problems.