‹ PublicAI Index
The LLM benchmark aggregator.
Kimi K2.7 Code
Moonshot · 1T · open weights
Strongest in Language (#36 of 62), weakest in Jailbreak resistance (#218 of 268). Above par in 5 of 16 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5 and GPT-6 Astra and ahead of GPT-4.1 mini and Grok 3.
§ 1 · Profile
What it is good at
Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.
Coding48.1−18.1#99/1792/6
Agentic coding48−18.9#96/1732/5
Agents50.8−15.6#117/2982/6
Knowledge work51.2−19.5#89/2082/2
Reasoning45−20.4#142/1952/5
Reasoning52.3−16.7#55/1471/3
Mathematics37.3−26.5#130/1521/2
Core abilities42.4−22.9#171/1991/4
Language49.8−16.8#36/621/1
Instruction following29.3−44.5#58/621/1
Safety49−12.4#206/3291/3
Secure code62.5−3.7#37/2701/1
Toxicity avoidance54.8−4.2#110/2681/1
Fairness45.9−24.1#172/2921/2
Harm refusal46.4−15.1#212/2921/2
Jailbreak resistance39.5−25.5#218/2681/1
§ 2 · Head to head
What it beats, and what beats it
The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Kimi K2.7 Code, left for the other.
§ 3 · Sources
Where the numbers come from
6 publications, 20 figures. Every one links to the page it was read from.
GDPval-AA 1046
AA-Briefcase 858
LiveBench 1.0 68.4LiveBench 1.0 · Reasoning 82.8LiveBench 1.0 · Coding 74LiveBench 1.0 · Agentic Coding 45.7LiveBench 1.0 · Mathematics 79.6LiveBench 1.0 · Data Analysis 62.7LiveBench 1.0 · Language 77.9LiveBench 1.0 · Instruction Following 56.3
SimpleBench 57.9%
Vals · SWE-bench Verified 78.2% (Mini-SWE-agent)Vals · Vibe Code Bench 47.21% (OpenHands)Vals · Code Migration 25.39%
Enkrypt · Jailbreak risk 21.6%Enkrypt · Harmful content risk 1.7%Enkrypt · CBRN risk 42%Enkrypt · Toxicity risk 2.9%Enkrypt · Bias risk 84.2%Enkrypt · Insecure code risk 7.6%
Badge
[](https://publicai.io/model-index/m/kimi-k2-7-code)