Claude Sonnet 4.5 vs Kimi K2.5
Wins 0 of 8 areas
—
Wins 7 of 8 areas
Coding · Agents · Reasoning · Images and charts · Math · Long documents · Following instructions
Kimi K2.5 is the stronger all-rounder.
Scores updated · 55 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 software15Kimi K2.55 of 8 tests · 2 ties
- Images and chartsUnderstanding pictures, charts and video03Kimi K2.53 of 3 tests
- ReasoningHard problems that need careful thinking13Kimi K2.53 of 4 tests
- Following instructionsDoing exactly what it is asked02Kimi K2.52 of 2 tests
- AgentsCarrying out multi-step tasks on its own23Kimi K2.53 of 5 tests
- Long documentsFinding answers in very long texts01Kimi K2.51 of 2 tests · 1 tie
- MathCompetition and research-level math01Kimi K2.51 of 1 test
- FactsGetting facts right instead of making them up22Even2 each
Math 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 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 Kimi K2.5 pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+50.8points ahead
- Recent programming contest problemsLiveCodeBench v6+21points ahead
- Multi-step tasks using many real software toolsMCP Atlas+20points ahead
Where Claude Sonnet 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+16.6points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+10.3points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+10.1points ahead
Every test, side by side
All 55 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.5
- LiveCodeBench v6Kimi K2.5 by 216485+21
- Terminal-Bench 2.1Claude Sonnet 4.5 by 10.155.845.7+10.1
- SWE-bench ProKimi K2.5 by 7.143.650.7+7.1
- SWE-bench MultilingualKimi K2.5 by 66773+6
- LMArena · WebDevKimi K2.5 by 42 rating points13941436+42 rating
- SciCodeKimi K2.5 by 3.345.749+3.3
- Terminal-Bench Hardtie35.634.8tie
- SWE-bench Verifiedtie77.276.8tie
AgentsKimi K2.5
- BrowseCompKimi K2.5 by 50.824.174.9+50.8
- MCP AtlasKimi K2.5 by 2043.863.8+20
- τ-Bench V3 · BankingClaude Sonnet 4.5 by 10.324.514.2+10.3
- GDPValClaude Sonnet 4.5 by 3.320.517.2+3.3
- OSWorld-VerifiedKimi K2.5 by 1.961.463.3+1.9
ReasoningKimi K2.5
- Humanity's Last ExamKimi K2.5 by 12.917.830.7+12.9
- GPQA DiamondKimi K2.5 by 4.583.487.9+4.5
- CritPtKimi K2.5 by 21.13.1+2
- ARC-AGI-2Claude Sonnet 4.5 by 1.813.611.8+1.8
FactsEven
- AA-Omniscience · Non-hallucinationClaude Sonnet 4.5 by 16.650.934.3+16.6
- SimpleQA VerifiedKimi K2.5 by 3.630.734.3+3.6
- AA-Omniscience · AccuracyKimi K2.5 by 2.332.935.2+2.3
- Vectara HHEM hallucination ratelower is betterClaude Sonnet 4.5 by 2.21214.2+2.2
Images and chartsKimi K2.5
- CharXiv (RQ)Kimi K2.5 by 10.367.277.5+10.3
- MathVistaKimi K2.5 by 10.379.890.1+10.3
- MMMU-ProKimi K2.5 by 6.768.775.4+6.7
Long documentsKimi K2.5
- AA-LCRKimi K2.5 by 5.772.378+5.7
- LongBench v2tie61.861tie
Following instructionsKimi K2.5
- IFBenchKimi K2.5 by 12.957.370.2+12.9
- Multi-ChallengeKimi K2.5 by 6.155.361.4+6.1
Other results26 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.
- HMMT Feb. 2025Kimi K2.5 by 27.967.595.4+27.9
- OJ-Bench (cpp)Kimi K2.5 by 2730.457.4+27
- HMMT 2025Kimi K2.5 by 20.874.695.4+20.8
- BrowseComp-ZHKimi K2.5 by 19.942.462.3+19.9
- HLE (with tools)Kimi K2.5 by 18.23250.2+18.2
- τ²-Bench Telecom (AA run)Kimi K2.5 by 17.878.195.9+17.8
- SimpleVQAKimi K2.5 by 13.657.671.2+13.6
- ToolathlonClaude Sonnet 4.5 by 13.24127.8+13.2
- τ²-BenchClaude Sonnet 4.5 by 12.69885.4+12.6
- xbench-DeepSearchKimi K2.5 by 10.76676.7+10.7
- AIME 2025Kimi K2.5 by 9.18796.1+9.1
- HMMT Nov. 2025Kimi K2.5 by 7.581.789.2+7.5
- AA-OmniscienceClaude Sonnet 4.5 by 7.2-0.1-7.3+7.2
- Artificial Analysis Coding IndexClaude Sonnet 4.5 by 5.352.146.8+5.3
- matharena_visual_math_overallKimi K2.5 by 4.875.880.6+4.8
- AA Agentic IndexKimi K2.5 by 4.217.521.7+4.2
- vectara_answer_rateClaude Sonnet 4.5 by 3.495.692.2+3.4
- OmniDocBench 1.5Kimi K2.5 by 385.888.8+3
- AA IntelligenceKimi K2.5 by 2.820.723.5+2.8
- vectara_factual_consistencyClaude Sonnet 4.5 by 2.28885.8+2.2
- ARC-AGI-1Kimi K2.5 by 1.663.765.3+1.6
- vectara_avg_summary_lengthClaude Sonnet 4.5 by 15.8 rating points127.8112+15.8 rating
- Terminal-Bench 2.0tie5050.8tie
- swe_bench_bashtie71.470.8tie
- MMLU-Protie87.587.1tie
- frontiermath_tier_4_v1tie4.24.2tie
Questions people ask
Which is better, Claude Sonnet 4.5 or Kimi K2.5?
Kimi K2.5 wins seven of the eight areas we test: coding, agents, reasoning, images and charts, math, long documents and following instructions. Claude Sonnet 4.5 wins none. They are level on facts.
Which is better for coding?
Kimi K2.5. It wins 5 of the 8 coding tests both models report; Claude Sonnet 4.5 wins 1, and 2 are ties.
Which is cheaper?
Claude Sonnet 4.5 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 55 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 26 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.