OPEN LINK ↗
// 1h agoBENCHMARK RESULT
Kimi K3.1 Pro Nears Opus on Terminal-Bench
Moonshot’s Kimi K3.1 Pro scores roughly 67% on Artificial Analysis’ Terminal-Bench 4.0 tracking, while Moonshot reports 65.7%—just behind Claude Opus 5.5 at 66.4%. The model is available now, with full open weights promised for November 4.
// ANALYSIS
This is a striking coding-agent leap, but the benchmark gap is narrow and the evaluation setup matters.
- –K3.1 Pro’s result represents a major improvement over K3’s reported 13%.
- –Moonshot evaluated it through the Claude Code harness, so results are not purely model-only comparisons.
- –The model offers a 1-million-token context window and a faster Highspeed variant.
- –Open weights could make frontier-level coding performance accessible for self-hosted deployments.
- –The official comparison is against Claude Opus 5.5—not Sonnet 5.5—and Moonshot’s own score remains slightly below it.
// TAGS
kimi-k3.1-prollmopen-weightscoding-agentai-codingbenchmarkevaluation
DISCOVERED
1h ago
2026-10-25
PUBLISHED
1h ago
2026-10-25
RELEVANCE
9/ 10
AUTHOR
lnbodicherla