Benchmark

Moonshot Kimi — reported benchmarks

Moonshot AI · source page ↗ · last checked Aug 22, 2026, 12:04 PM

Reported benchmarks · Kimi K3

captured Aug 22, 2026, 12:04 PM
BenchmarkScore
GPQA Diamond93.5%
SWE-Bench Verified76.8%
Terminal-Bench 2.188.3
FrontierSWE81.2 pass@1
Program Bench77.8% · raw pass rate
SWE Marathon42.0
DeepSWE67.5 · Kimi Code harness
BrowseComp91.2 · 300K context compaction
Humanity's Last Exam (HLE-Full)43.5 / 56.0 · text / text+tools
CritPt23.4
AA-LCR74.7 · Long-context retrieval

Vendor-reported via automated web search — not independently verified. See the cited matrix on /models.

Change history