Moonshot AI20 Apr 2026Model

Kimi K2.6

Benchmark results

Benchmark	Score	Source check
Humanity's Last ExamNo tools · full set (36.4 text-only); up to 98,304 output tokens	34.7%	Checked A person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Humanity's Last ExamWith tools · full set (55.5 text-only); search, code and browsing tools; context management	54.0%	Checked A person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
SWE-bench Proin-house harness adapted from SWE-agent; thinking mode; average of 10 runs · The source doesn't say whether this is the public or the commercial set	58.6%	Checked A person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. Corroborated The same number is printed in 3 other lab documents: 1 2 3
GPQA Diamondthinking mode; up to 98,304 output tokens	90.5%	Checked A person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
BrowseCompwith context management, discard-all strategy	83.2%	Checked A person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. Corroborated The same number is printed in 3 other lab documents: 1 2 3

Sources

Reported by Moonshot AI

Kimi K2.6: Advancing Open-Source Codingkimi.ai · 20 Apr 2026
Our capture of the source
Our capture of the source