Claude Opus 4.6 vs Kimi K2.6
Wins 3 of 8 areas
Coding · Reasoning · Math
Wins 3 of 8 areas
Images and charts · Long documents · Following instructions
The two are evenly matched.Claude Opus 4.6 is better at math and coding; Kimi K2.6 at following instructions and images and charts.
Scores updated · 78 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- MathCompetition and research-level math51Claude Opus 4.65 of 7 tests · 1 tie
- CodingWriting and fixing software41Claude Opus 4.64 of 9 tests · 4 ties
- ReasoningHard problems that need careful thinking31Claude Opus 4.63 of 4 tests
- Following instructionsDoing exactly what it is asked02Kimi K2.62 of 2 tests
- Images and chartsUnderstanding pictures, charts and video12Kimi K2.62 of 3 tests
- Long documentsFinding answers in very long texts01Kimi K2.61 of 1 test
- AgentsCarrying out multi-step tasks on its own22Even2 each · 1 tie
- FactsGetting facts right instead of making them up22Even2 each
Long documents rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Kimi K2.6 costs 84% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Claude Opus 4.6 pulls ahead
- Real work tasks from 44 professionsGDPVal+28.9points ahead
- Proof problems from the US Math OlympiadUSAMO 2026+15points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+14.4points ahead
Where Kimi K2.6 pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+22.9points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+22.3points ahead
- Olympiad math problems with short answersIMOAnswerBench+10.7points ahead
Every test, side by side
All 78 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.6
- SWE-bench ProKimi K2.6 by 5.253.458.6+5.2
- LMArena · WebDevClaude Opus 4.6 by 37 rating points15461509+37 rating
- Terminal-Bench HardClaude Opus 4.6 by 2.346.243.9+2.3
- LiveBench · Agentic CodingClaude Opus 4.6 by 2.14946.9+2.1
- SWE-bench MultilingualClaude Opus 4.6 by 1.177.876.7+1.1
- LiveCodeBench v6tie88.889.6tie
- SWE-bench Verifiedtie80.880.2tie
- LiveBench · Codingtie78.278.6tie
- SciCodetie51.951.5tie
AgentsEven
- GDPValClaude Opus 4.6 by 28.955.927+28.9
- MCP AtlasKimi K2.6 by 5.462.768.1+5.4
- AA ApexAgentsClaude Opus 4.6 by 4.53328.5+4.5
- BrowseCompKimi K2.6 by 2.38486.3+2.3
- OSWorld-Verifiedtie72.773.1tie
ReasoningClaude Opus 4.6
- LiveBench · ReasoningClaude Opus 4.6 by 9.388.779.4+9.3
- CritPtClaude Opus 4.6 by 4.612.68+4.6
- Humanity's Last ExamClaude Opus 4.6 by 2.439.937.5+2.4
- GPQA DiamondKimi K2.6 by 1.589.691.1+1.5
FactsEven
- AA-Omniscience · Non-hallucinationKimi K2.6 by 22.337.259.5+22.3
- AA-Omniscience · AccuracyClaude Opus 4.6 by 14.44732.6+14.4
- SimpleQA VerifiedClaude Opus 4.6 by 12.14734.9+12.1
- Vectara HHEM hallucination ratelower is betterKimi K2.6 by 1.412.210.8+1.4
Images and chartsKimi K2.6
- CharXiv (RQ)Kimi K2.6 by 9.377.486.7+9.3
- MMMU-ProKimi K2.6 by 475.479.4+4
- LMArena · VisionClaude Opus 4.6 by 34 rating points13161282+34 rating
MathClaude Opus 4.6
- USAMO 2026Claude Opus 4.6 by 1566.251.2+15
- IMOAnswerBenchKimi K2.6 by 10.775.386+10.7
- FrontierMath Tiers 1-3 (v2)Claude Opus 4.6 by 8.86657.2+8.8
- LiveBench · MathematicsClaude Opus 4.6 by 589.384.3+5
- HMMT Feb. 2026Claude Opus 4.6 by 3.596.292.7+3.5
- FrontierMath Tier 4Claude Opus 4.6 by 1.226.825.6+1.2
- AIME 2026tie96.796.4tie
Following instructionsKimi K2.6
- IFBenchKimi K2.6 by 22.953.176+22.9
- LiveBench · Instruction FollowingKimi K2.6 by 1.163.364.4+1.1
Other results43 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- BabyVisionKimi K2.6 by 53.714.868.5+53.7
- AA Agentic IndexClaude Opus 4.6 by 45.567.622.1+45.5
- baby_vision_with_pythonKimi K2.6 by 30.138.468.5+30.1
- MATH-VisionKimi K2.6 by 27.765.593.2+27.7
- HiL-Bench (Tools-allowed)Claude Opus 4.6 by 19.638.318.7+19.6
- Artificial Analysis Coding IndexKimi K2.6 by 13.748.161.8+13.7
- GDPval-AA (Elo)Claude Opus 4.6 by 137 rating points16191482+137 rating
- Apex (Pass@1)Claude Opus 4.6 by 10.534.524+10.5
- V* (w/ python)Kimi K2.6 by 10.586.496.9+10.5
- Apex Shortlist (Pass@1)Claude Opus 4.6 by 10.485.975.5+10.4
- AgentWorldBench - OSClaude Opus 4.6 by 9.470.260.8+9.4
- mathvision_with_pythonKimi K2.6 by 8.684.693.2+8.6
- AA-OmniscienceClaude Opus 4.6 by 8.413.75.3+8.4
- DeepSearchQA (F1)Claude Opus 4.6 by 8.391.383+8.3
- frontiermath_tier_4_v1Claude Opus 4.6 by 8.322.914.6+8.3
- livebench_languageClaude Opus 4.6 by 8.283.375.1+8.2
- MCPAtlas Public (Pass@1)Claude Opus 4.6 by 7.273.866.6+7.2
- ToolathlonClaude Opus 4.6 by 6.856.850+6.8
- AgentWorldBench - SWEClaude Opus 4.6 by 5.864.558.8+5.8
- AgentWorldBench - TerminalClaude Opus 4.6 by 557.552.5+5
- AA IntelligenceClaude Opus 4.6 by 4.931.927+4.9
- livebench_data_analysisClaude Opus 4.6 by 4.869.965.1+4.8
- AgentWorldBench - MCPClaude Opus 4.6 by 4.769.965.2+4.7
- AgentWorldBench - OverallClaude Opus 4.6 by 4.457.853.4+4.4
- LiveBenchClaude Opus 4.6 by 4.176.372.2+4.1
- τ²-Bench Telecom (AA run)Kimi K2.6 by 3.892.195.9+3.8
- AgentWorldBench - AndroidClaude Opus 4.6 by 2.861.758.9+2.8
- mmmu_pro_with_pythonKimi K2.6 by 2.877.380.1+2.8
- BrowseComp (Agent Swarm)Kimi K2.6 by 2.683.786.3+2.6
- DeepSearchQA (accuracy)Kimi K2.6 by 2.480.683+2.4
- vectara_avg_summary_lengthClaude Opus 4.6 by 20.9 rating points137.6116.7+20.9 rating
- charxiv_rq_with_pythonKimi K2.6 by 284.786.7+2
- MMLU-ProClaude Opus 4.6 by 289.187.1+2
- AgentWorldBench - SearchClaude Opus 4.6 by 1.829.327.5+1.8
- vectara_factual_consistencyKimi K2.6 by 1.487.889.2+1.4
- Terminal-Bench 2.0Kimi K2.6 by 1.365.466.7+1.3
- AgentWorldBench - WebClaude Opus 4.6 by 1.251.450.2+1.2
- LiveCodeBenchtie88.889.6tie
- MCPMarktie56.755.9tie
- HLE (with tools)tie53.354tie
- Chinese SimpleQA (C-SimpleQA)tie76.475.9tie
- OJBench (python)tie60.360.6tie
- vectara_answer_ratetie99.899.7tie
Questions people ask
Which is better, Claude Opus 4.6 or Kimi K2.6?
Claude Opus 4.6 and Kimi K2.6 each win three of the eight areas we test. Claude Opus 4.6 wins coding, reasoning and math; Kimi K2.6 wins images and charts, long documents and following instructions. Kimi K2.6 costs 84% less. They are level on agents and facts.
Which is better for coding?
Claude Opus 4.6. It wins 4 of the 9 coding tests both models report; Kimi K2.6 wins 1, and 4 are ties.
Which is cheaper?
Claude Opus 4.6 costs $5.00 per million input tokens and $25.00 per million output tokens; Kimi K2.6 costs $0.95 and $4.00. That makes Kimi K2.6 about 84% cheaper for the same work.
How do you compare the two?
We use the 78 benchmark tests both models have published scores on. The verdict counts the 35 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 43 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.