Claude Sonnet 4.6 vs Kimi K2.5
Wins 5 of 7 areas
Coding · Agents · Reasoning · Facts · Long documents
Wins 2 of 7 areas
Images and charts · Following instructions
Claude Sonnet 4.6 is the stronger all-rounder.Kimi K2.5 is cheaper and better at images and charts.
Scores updated · 40 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. 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 software60Claude Sonnet 4.66 of 6 tests
- AgentsCarrying out multi-step tasks on its own41Claude Sonnet 4.64 of 6 tests · 1 tie
- ReasoningHard problems that need careful thinking20Claude Sonnet 4.62 of 4 tests · 2 ties
- FactsGetting facts right instead of making them up31Claude Sonnet 4.63 of 4 tests
- Long documentsFinding answers in very long texts10Claude Sonnet 4.61 of 1 test
- Images and chartsUnderstanding pictures, charts and video12Kimi K2.52 of 3 tests
- Following instructionsDoing exactly what it is asked01Kimi K2.51 of 1 test
Long documents and following instructions rest on a single test each.
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 80% 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 Sonnet 4.6 pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+46.5points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+25.5points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+20.2points ahead
Where Kimi K2.5 pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+13.6points ahead
- Reasoning about charts from research papersCharXiv (RQ)+5.9points ahead
- Multi-step tasks using many real software toolsMCP Atlas+2.5points ahead
Every test, side by side
All 40 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 4.6
- Terminal-Bench 2.1Claude Sonnet 4.6 by 25.571.245.7+25.5
- Terminal-Bench HardClaude Sonnet 4.6 by 18.25334.8+18.2
- LMArena · WebDevClaude Sonnet 4.6 by 86 rating points15221436+86 rating
- SWE-bench ProClaude Sonnet 4.6 by 7.458.150.7+7.4
- SWE-bench VerifiedClaude Sonnet 4.6 by 2.879.676.8+2.8
- SciCodeClaude Sonnet 4.6 by 1.150.149+1.1
AgentsClaude Sonnet 4.6
- τ-Bench V3 · BankingClaude Sonnet 4.6 by 20.234.414.2+20.2
- GDPValClaude Sonnet 4.6 by 19.536.717.2+19.5
- AA ApexAgentsClaude Sonnet 4.6 by 16.52811.5+16.5
- OSWorld-VerifiedClaude Sonnet 4.6 by 9.272.563.3+9.2
- MCP AtlasKimi K2.5 by 2.561.363.8+2.5
- BrowseComptie74.774.9tie
ReasoningClaude Sonnet 4.6
- ARC-AGI-2Claude Sonnet 4.6 by 46.558.311.8+46.5
- Humanity's Last ExamClaude Sonnet 4.6 by 2.933.630.7+2.9
- GPQA Diamondtie87.587.9tie
- CritPttie3.13.1tie
FactsClaude Sonnet 4.6
- AA-Omniscience · Non-hallucinationClaude Sonnet 4.6 by 17.351.634.3+17.3
- AA-Omniscience · AccuracyClaude Sonnet 4.6 by 5.740.935.2+5.7
- Vectara HHEM hallucination ratelower is betterClaude Sonnet 4.6 by 3.610.614.2+3.6
- SimpleQA VerifiedKimi K2.5 by 1.532.834.3+1.5
Images and chartsKimi K2.5
- CharXiv (RQ)Kimi K2.5 by 5.971.677.5+5.9
- MMMU-ProKimi K2.5 by 2.173.375.4+2.1
- LMArena · VisionClaude Sonnet 4.6 by 14 rating points12831269+14 rating
Following instructionsKimi K2.5
- IFBenchKimi K2.5 by 13.656.670.2+13.6
Other results15 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.
- GDPval-AA v2Claude Sonnet 4.6 by 386 rating points13951009+386 rating
- ARC-AGI-1Claude Sonnet 4.6 by 20.78665.3+20.7
- τ²-Bench Telecom (AA run)Kimi K2.5 by 20.275.795.9+20.2
- AA-OmniscienceClaude Sonnet 4.6 by 19.512.2-7.3+19.5
- Artificial Analysis Coding IndexClaude Sonnet 4.6 by 16.26346.8+16.2
- ToolathlonClaude Sonnet 4.6 by 13.24127.8+13.2
- AA Agentic IndexClaude Sonnet 4.6 by 11.433.121.7+11.4
- Terminal-Bench 2.0Claude Sonnet 4.6 by 8.359.150.8+8.3
- vectara_answer_rateClaude Sonnet 4.6 by 7.799.992.2+7.7
- AA IntelligenceClaude Sonnet 4.6 by 6.630.123.5+6.6
- LiveBenchClaude Sonnet 4.6 by 6.475.569.1+6.4
- frontiermath_tier_4_v1Claude Sonnet 4.6 by 4.18.34.2+4.1
- vectara_factual_consistencyClaude Sonnet 4.6 by 3.689.485.8+3.6
- HLE (with tools)Kimi K2.5 by 3.446.850.2+3.4
- vectara_avg_summary_lengthtie114.7112tie
Questions people ask
Which is better, Claude Sonnet 4.6 or Kimi K2.5?
Claude Sonnet 4.6 wins five of the seven areas where both have results: coding, agents, reasoning, facts and long documents. Kimi K2.5 wins images and charts and following instructions, and costs 80% less.
Which is better for coding?
Claude Sonnet 4.6. It wins 6 of the 6 coding tests both models report; Kimi K2.5 wins none.
Which is cheaper?
Claude Sonnet 4.6 costs $3.00 per million input tokens and $15.00 per million output tokens; Kimi K2.5 costs $0.60 and $3.00. That makes Kimi K2.5 about 80% cheaper for the same work.
How do you compare the two?
We use the 40 benchmark tests both models have published scores on. The verdict counts the 25 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 15 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.