Benchmark: Kimi K3 outperforms Codex in kernel optimization task

Xianbao_QIAN · x · 2026-08-26

A developer shared Kimi K3's performance on Triton kernel writing tasks. By pitting Kimi K3 (high mode) against Codex (Sol medium) on the same problem, results showed that the kernels generated by Kimi K3 were significantly better than Codex's approach. The test was run locally via vLLM.

Original post →

More from Models

Models channel →