#benchmarks

Open Source Models Are Beating Frontier AI at Code

Kimi K2.6 outperforms GPT-5.4 on SWE-Bench Pro. Costs 25x less. Nobody's switching.