Claude Sonnet 4 vs Kimi K2.5
Wins 1 of 7 areas
Facts
Wins 6 of 7 areas
Coding · Agents · Reasoning · Images and charts · Long documents · Following instructions
Kimi K2.5 is the stronger all-rounder.Claude Sonnet 4 is better at facts.
Scores updated · 41 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 software07Kimi K2.57 of 7 tests
- ReasoningHard problems that need careful thinking04Kimi K2.54 of 4 tests
- AgentsCarrying out multi-step tasks on its own13Kimi K2.53 of 4 tests
- Images and chartsUnderstanding pictures, charts and video02Kimi K2.52 of 2 tests
- Following instructionsDoing exactly what it is asked02Kimi K2.52 of 2 tests
- Long documentsFinding answers in very long texts01Kimi K2.51 of 1 test
- FactsGetting facts right instead of making them up21Claude Sonnet 42 of 3 tests
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 80% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Kimi K2.5 pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+62.7points ahead
- Recent programming contest problemsLiveCodeBench v6+36.5points ahead
- Completes tasks by operating a computer desktopOSWorld-Verified+21.1points ahead
Where Claude Sonnet 4 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+36.6points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+2.5points ahead
Every test, side by side
All 41 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 36.548.585+36.5
- SWE-bench MultilingualKimi K2.5 by 16.156.973+16.1
- Terminal-Bench 2.1Kimi K2.5 by 9.436.345.7+9.4
- SciCodeKimi K2.5 by 94049+9
- SWE-bench ProKimi K2.5 by 842.750.7+8
- SWE-bench VerifiedKimi K2.5 by 4.172.776.8+4.1
- Terminal-Bench HardKimi K2.5 by 3.731.134.8+3.7
AgentsKimi K2.5
- BrowseCompKimi K2.5 by 62.712.274.9+62.7
- OSWorld-VerifiedKimi K2.5 by 21.142.263.3+21.1
- GDPValKimi K2.5 by 7.59.717.2+7.5
- τ-Bench V3 · BankingClaude Sonnet 4 by 2.516.714.2+2.5
ReasoningKimi K2.5
- Humanity's Last ExamKimi K2.5 by 2010.730.7+20
- GPQA DiamondKimi K2.5 by 10.277.787.9+10.2
- ARC-AGI-2Kimi K2.5 by 5.95.911.8+5.9
- CritPtKimi K2.5 by 2.80.33.1+2.8
FactsClaude Sonnet 4
- AA-Omniscience · Non-hallucinationClaude Sonnet 4 by 36.670.934.3+36.6
- AA-Omniscience · AccuracyKimi K2.5 by 12.522.735.2+12.5
- Vectara HHEM hallucination ratelower is betterClaude Sonnet 4 by 3.910.314.2+3.9
Images and chartsKimi K2.5
- MMMU-ProKimi K2.5 by 13.661.875.4+13.6
- LMArena · VisionKimi K2.5 by 78 rating points11911269+78 rating
Following instructionsKimi K2.5
- IFBenchKimi K2.5 by 15.554.770.2+15.5
- Multi-ChallengeKimi K2.5 by 4.357.161.4+4.3
Other results18 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 2025Kimi K2.5 by 79.515.995.4+79.5
- τ²-BenchKimi K2.5 by 40.245.285.4+40.2
- BrowseComp-ZHKimi K2.5 by 33.229.162.3+33.2
- τ²-Bench Telecom (AA run)Kimi K2.5 by 31.364.695.9+31.3
- HLE (with tools)Kimi K2.5 by 29.920.350.2+29.9
- AIME 2025Kimi K2.5 by 25.670.596.1+25.6
- ARC-AGI-1Kimi K2.5 by 25.34065.3+25.3
- xbench-DeepSearchKimi K2.5 by 12.164.676.7+12.1
- Artificial Analysis Coding IndexKimi K2.5 by 9.237.646.8+9.2
- AA-OmniscienceClaude Sonnet 4 by 7.50.2-7.3+7.5
- vectara_answer_rateClaude Sonnet 4 by 6.498.692.2+6.4
- LiveBenchClaude Sonnet 4 by 5.774.869.1+5.7
- AA IntelligenceKimi K2.5 by 4.618.923.5+4.6
- frontiermath_tier_4_v1Kimi K2.5 by 4.204.2+4.2
- AA Agentic IndexKimi K2.5 by 4.117.621.7+4.1
- vectara_factual_consistencyClaude Sonnet 4 by 3.989.785.8+3.9
- MMLU-ProKimi K2.5 by 3.483.787.1+3.4
- vectara_avg_summary_lengthClaude Sonnet 4 by 33.8 rating points145.8112+33.8 rating
Questions people ask
Which is better, Claude Sonnet 4 or Kimi K2.5?
Kimi K2.5 wins six of the seven areas where both have results: coding, agents, reasoning, images and charts, long documents and following instructions. Claude Sonnet 4 wins facts.
Which is better for coding?
Kimi K2.5. It wins 7 of the 7 coding tests both models report; Claude Sonnet 4 wins none.
Which is cheaper?
Claude Sonnet 4 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 41 benchmark tests both models have published scores on. The verdict counts the 23 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 18 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.