Kimi K2.6

Kimi K2.6 is a reasoning model from Moonshot AI. 34 benchmarks count toward its score, in 7 categories.

availableShows if the model has enough results for an index.
IndexOverall score. 50 is the middle.60.9 ±3.3
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.38/s
Input / 1MUS dollars per 1M input tokens.$0.95
Output / 1MUS dollars per 1M output tokens.$4
ContextMaximum tokens in one request.256K
EloLMArena rating and rank.N/A

50 is the middle of the board. The range shows the doubt in the index.

CapabilitiesScore per category. 50 is the middle.

50 is the middle
AgenticMulti-step tasks with tools.
61.9
CodingCode writing and repair.
61.8
ReasoningLogic problems and puzzles.
59.7
MultimodalTasks with images and text.
59.0
KnowledgeFacts and expert knowledge.
59.6
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
66.2
MathMath problems.
57.4

Results

34 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result on the index scale.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
V*Multimodal96.9%57.6Z.AI
AIME 2026Math96.4%55.3Qwen
τ²-Bench Tool-Agent-User EvaluationAgentic95.9%67.6Victor Barres et al.
Harvard-MIT Mathematics Tournament February 2026Math92.7%59.5Qwen
DeepSearchQAAgentic92.5%71.6Meta AI
Artificial Analysis GPQA DiamondKnowledge91.1%61.9Artificial Analysis
Graduate-Level Google-Proof Q&AKnowledge90.5%61.8David Rein et al.
GPQA DiamondKnowledge90.5%61.8David Rein et al.
LiveCodeBench v6Coding89.6%58.6LiveCodeBench maintainers
MathVisionMultimodal87.4%Qwen
MMAnswerBenchMath86.0%Qwen
BrowseCompAgentic83.2%69.6OpenAI
Artificial Analysis Long Context ReasoningReasoning81.0%64.3Artificial Analysis
WideResearchAgentic80.8%66.7Qwen
CharXiv ReasoningMultimodal80.4%57.6CharXiv authors
Software Engineering Benchmark VerifiedCoding80.2%61.9Carlos E. Jimenez et al.
MMMU-Pro with PythonMultimodal80.1%OpenAI
Massive Multi-discipline Multimodal Understanding ProMultimodal79.4%56.6MMMU-Pro authors
Artificial Analysis MMMU-ProMultimodal79.4%64.1Artificial Analysis
Artificial Analysis IFBenchInstruction76.0%66.2Artificial Analysis
OSWorld-VerifiedAgentic73.1%61.8Tianbao Xie et al.
Claw-EvalAgentic62.3%56.3Bowen Ye et al.
Artificial Analysis Coding IndexCoding61.8%62.5Artificial Analysis
SWE-bench ProCoding58.6%60.7Xiang Deng et al.
Gert Labs Composite Game BenchmarkAgentic56.8%62.9Gert Labs
MCP AtlasAgentic55.9%50.3OpenAI
Scientific Code BenchmarkCoding52.2%62.9Benchmark authors
Artificial Analysis SciCodeCoding51.5%64.3Artificial Analysis
ToolathlonAgentic50.0%64.6OpenAI
cursorBench31Coding47.6%Benchmark authors
Artificial Analysis Humanity's Last ExamKnowledge37.5%65.2Artificial Analysis
Humanity's Last ExamKnowledge34.7%58.2Center for AI Safety et al.
Artificial Analysis Omniscience AccuracyKnowledge32.6%54.2Artificial Analysis
APEX-Agents-AAAgentic28.5%63.6Artificial Analysis / Mercor
Artificial Analysis Intelligence IndexKnowledge27.0%56.1Artificial Analysis
GDPval-AA normalizedAgentic26.3%55.8Artificial Analysis
Artificial Analysis Agentic IndexAgentic22.1%55.2Artificial Analysis
ResearchClawBenchAgentic18.0%InternScience
Critical Physics TasksReasoning8.0%55.2Artificial Analysis
OSWorld 2.0Agentic4.6%58.2Mengqi Yuan et al.

34 benchmarks count, from 35 of 40 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authors

Same level, lower price

Qwen3.8-Flash-Next66.4 · Freedots3-note Preview65.5 · FreeApodex 1.1 Mini59.1 · Free

More from Moonshot AI

Kimi K372.0Kimi K2.7 Code59.5Kimi K2.5 (Reasoning)54.6Kimi K2.553.3Kimi K2.5 Thinking53.8