# @witcheer — 2026-04-20

♥9 ↻0 · https://x.com/witcheer/status/2046321949476102149

kimi k2.6 is open-weight and beat GPT-5.4 on SWE-Bench Pro (58.6 vs 57.7)

k2.6 sustained 4,000+ tool calls over 12 hours to optimise a zig inference engine from 15 to 193 tokens/sec. most models degrade after a few hundred calls. this one iterated through 14 rounds of profiling

tags: author:witcheer, capability-shock, kind:tweet, model:kimi-k2, on:kimi-k2-6, year:2026
cited on: kimi-k2-6