Claude Opus 4.6 vs Kimi K2.5
Wins 5 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts
Wins 1 of 8 areas
Following instructions
Claude Opus 4.6 is the stronger all-rounder.Kimi K2.5 is cheaper and better at following instructions.
Scores updated · 65 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.
- CodingWriting and fixing software70Claude Opus 4.67 of 7 tests
- ReasoningHard problems that need careful thinking40Claude Opus 4.64 of 4 tests
- FactsGetting facts right instead of making them up40Claude Opus 4.64 of 4 tests
- AgentsCarrying out multi-step tasks on its own41Claude Opus 4.64 of 5 tests
- Images and chartsUnderstanding pictures, charts and video10Claude Opus 4.61 of 3 tests · 2 ties
- Following instructionsDoing exactly what it is asked02Kimi K2.52 of 2 tests
- MathCompetition and research-level math11Even1 each · 1 tie
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
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 K2.5 costs 88% 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
- Abstract visual puzzles that people can solveARC-AGI-2+57points ahead
- Real work tasks from 44 professionsGDPVal+38.7points ahead
- Long professional tasks in banking, consulting and lawAA ApexAgents+21.5points ahead
Where Kimi K2.5 pulls ahead
- Keeps track of context across a multi-turn chatMulti-Challenge+24.2points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+17.1points ahead
- Olympiad math problems with short answersIMOAnswerBench+6.5points ahead
Every test, side by side
All 65 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.6
- Terminal-Bench HardClaude Opus 4.6 by 11.446.234.8+11.4
- LMArena · WebDevClaude Opus 4.6 by 110 rating points15461436+110 rating
- SWE-bench MultilingualClaude Opus 4.6 by 4.877.873+4.8
- SWE-bench VerifiedClaude Opus 4.6 by 480.876.8+4
- LiveCodeBench v6Claude Opus 4.6 by 3.888.885+3.8
- SciCodeClaude Opus 4.6 by 2.951.949+2.9
- SWE-bench ProClaude Opus 4.6 by 2.753.450.7+2.7
AgentsClaude Opus 4.6
- GDPValClaude Opus 4.6 by 38.755.917.2+38.7
- AA ApexAgentsClaude Opus 4.6 by 21.53311.5+21.5
- OSWorld-VerifiedClaude Opus 4.6 by 9.472.763.3+9.4
- BrowseCompClaude Opus 4.6 by 9.18474.9+9.1
- MCP AtlasKimi K2.5 by 1.162.763.8+1.1
ReasoningClaude Opus 4.6
- ARC-AGI-2Claude Opus 4.6 by 5768.811.8+57
- CritPtClaude Opus 4.6 by 9.512.63.1+9.5
- Humanity's Last ExamClaude Opus 4.6 by 9.239.930.7+9.2
- GPQA DiamondClaude Opus 4.6 by 1.789.687.9+1.7
FactsClaude Opus 4.6
- SimpleQA VerifiedClaude Opus 4.6 by 12.74734.3+12.7
- AA-Omniscience · AccuracyClaude Opus 4.6 by 11.84735.2+11.8
- AA-Omniscience · Non-hallucinationClaude Opus 4.6 by 2.937.234.3+2.9
- Vectara HHEM hallucination ratelower is betterClaude Opus 4.6 by 212.214.2+2
Images and chartsClaude Opus 4.6
- LMArena · VisionClaude Opus 4.6 by 47 rating points13161269+47 rating
- CharXiv (RQ)tie77.477.5tie
- MMMU-Protie75.475.4tie
MathEven
- HMMT Feb. 2026Claude Opus 4.6 by 9.196.287.1+9.1
- IMOAnswerBenchKimi K2.5 by 6.575.381.8+6.5
- AIME 2026tie96.795.8tie
Following instructionsKimi K2.5
- Multi-ChallengeKimi K2.5 by 24.237.161.4+24.2
- IFBenchKimi K2.5 by 17.153.170.2+17.1
Other results36 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.
- AA Agentic IndexClaude Opus 4.6 by 45.967.621.7+45.9
- CyberGymClaude Opus 4.6 by 32.573.841.3+32.5
- ToolathlonClaude Opus 4.6 by 2956.827.8+29
- ARC-AGI-1Claude Opus 4.6 by 27.79365.3+27.7
- MCPMarkClaude Opus 4.6 by 27.256.729.5+27.2
- BabyVisionKimi K2.5 by 21.714.836.5+21.7
- AA-OmniscienceClaude Opus 4.6 by 2113.7-7.3+21
- Tool-DecathlonClaude Opus 4.6 by 19.447.227.8+19.4
- frontiermath_tier_4_v1Claude Opus 4.6 by 18.722.94.2+18.7
- MATH-VisionKimi K2.5 by 18.765.584.2+18.7
- NL2RepoClaude Opus 4.6 by 17.849.832+17.8
- Terminal-Bench 2.0Claude Opus 4.6 by 14.665.450.8+14.6
- DeepSearchQA (F1)Claude Opus 4.6 by 14.291.377.1+14.2
- LVBenchKimi K2.5 by 12.96375.9+12.9
- browsecomp_with_context_managerClaude Opus 4.6 by 9.18474.9+9.1
- AA IntelligenceClaude Opus 4.6 by 8.431.923.5+8.4
- matharena_visual_math_overallKimi K2.5 by 8.372.380.6+8.3
- vectara_answer_rateClaude Opus 4.6 by 7.699.892.2+7.6
- LiveBenchClaude Opus 4.6 by 7.276.369.1+7.2
- HMMT Nov. 2025Claude Opus 4.6 by 7.196.389.2+7.1
- τ³-BenchClaude Opus 4.6 by 6.472.466+6.4
- charxiv_rq_with_pythonClaude Opus 4.6 by 684.778.7+6
- OJBench (python)Claude Opus 4.6 by 5.660.354.7+5.6
- BrowseComp (Agent Swarm)Claude Opus 4.6 by 5.383.778.4+5.3
- τ²-Bench Telecom (AA run)Kimi K2.5 by 3.892.195.9+3.8
- AIME 2025Claude Opus 4.6 by 3.799.896.1+3.7
- DeepSearchQA (accuracy)Claude Opus 4.6 by 3.580.677.1+3.5
- HLE (with tools)Claude Opus 4.6 by 3.153.350.2+3.1
- vectara_avg_summary_lengthClaude Opus 4.6 by 25.6 rating points137.6112+25.6 rating
- baby_vision_with_pythonKimi K2.5 by 2.138.440.5+2.1
- MMLU-ProClaude Opus 4.6 by 289.187.1+2
- vectara_factual_consistencyClaude Opus 4.6 by 287.885.8+2
- Artificial Analysis Coding IndexClaude Opus 4.6 by 1.348.146.8+1.3
- V* (w/ python)tie86.486.9tie
- mathvision_with_pythontie84.685tie
- mmmu_pro_with_pythontie77.377.7tie
Questions people ask
Which is better, Claude Opus 4.6 or Kimi K2.5?
Claude Opus 4.6 wins five of the eight areas we test: coding, agents, reasoning, facts and images and charts. Kimi K2.5 wins following instructions, and costs 88% less. They are level on math and long documents.
Which is better for coding?
Claude Opus 4.6. It wins 7 of the 7 coding tests both models report; Kimi K2.5 wins none.
Which is cheaper?
Claude Opus 4.6 costs $5.00 per million input tokens and $25.00 per million output tokens; Kimi K2.5 costs $0.60 and $3.00. That makes Kimi K2.5 about 88% cheaper for the same work.
How do you compare the two?
We use the 65 benchmark tests both models have published scores on. The verdict counts the 29 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 36 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.