Kimi K3 scored 57 on the Artificial Analysis Coding Agent Index, matching GPT-5.6 Terra and GPT-5.5, while outperforming Opus 4.8. The model demonstrated strong performance across three coding evaluations, including 84% on Terminal-Bench v2 and 64% on DeepSWE. K3 is also cost-efficient, averaging $3.18 per task, making it significantly cheaper than several competitors like GPT-5.6 Sol max and Fable 5 max.
A new coding agent from Kimi offers competitive performance at a lower cost, increasing pressure on established frontier models to justify their pricing.