Kimi K2.5 vs Qwen3.5 35B A3B
Wins 7 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents
Wins 0 of 8 areas
—
Kimi K2.5 is the stronger all-rounder.Qwen3.5 35B A3B is cheaper.
Both rank among the ten best models we track in images and charts.
Scores updated · 58 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 software80Kimi K2.58 of 8 tests
- AgentsCarrying out multi-step tasks on its own50Kimi K2.55 of 5 tests
- Images and chartsUnderstanding pictures, charts and video30Kimi K2.53 of 4 tests · 1 tie
- ReasoningHard problems that need careful thinking30Kimi K2.53 of 3 tests
- MathCompetition and research-level math30Kimi K2.53 of 3 tests
- Long documentsFinding answers in very long texts20Kimi K2.52 of 2 tests
- FactsGetting facts right instead of making them up21Kimi K2.52 of 3 tests
- Following instructionsDoing exactly what it is asked11Even1 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.
Qwen3.5 35B A3B costs 38% 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
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+19.7points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+15.1points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+13.9points ahead
Where Qwen3.5 35B A3B pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+2.3points ahead
Every test, side by side
All 58 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.5
- LMArena · WebDevKimi K2.5 by 186 rating points14361250+186 rating
- SWE-bench MultilingualKimi K2.5 by 12.77360.3+12.7
- SciCodeKimi K2.5 by 11.34937.7+11.3
- LiveCodeBench v6Kimi K2.5 by 10.48574.6+10.4
- Terminal-Bench HardKimi K2.5 by 8.334.826.5+8.3
- SWE-bench VerifiedKimi K2.5 by 7.676.869.2+7.6
- SWE-bench ProKimi K2.5 by 6.150.744.6+6.1
- Terminal-Bench 2.1Kimi K2.5 by 4.945.740.8+4.9
AgentsKimi K2.5
- BrowseCompKimi K2.5 by 13.974.961+13.9
- GDPValKimi K2.5 by 11.617.25.6+11.6
- τ-Bench V3 · BankingKimi K2.5 by 9.314.24.9+9.3
- OSWorld-VerifiedKimi K2.5 by 8.863.354.5+8.8
- MCP AtlasKimi K2.5 by 1.463.862.4+1.4
ReasoningKimi K2.5
- Humanity's Last ExamKimi K2.5 by 9.730.721+9.7
- GPQA DiamondKimi K2.5 by 3.487.984.5+3.4
- CritPtKimi K2.5 by 2.23.10.9+2.2
FactsKimi K2.5
- AA-Omniscience · Non-hallucinationKimi K2.5 by 19.734.314.6+19.7
- AA-Omniscience · AccuracyKimi K2.5 by 15.135.220.1+15.1
- Vectara HHEM hallucination ratelower is betterQwen3.5 35B A3B by 3.714.210.5+3.7
Images and chartsKimi K2.5
- MathVistaKimi K2.5 by 3.990.186.2+3.9
- MMMU-ProKimi K2.5 by 2.775.472.7+2.7
- OCRBenchKimi K2.5 by 1.392.391+1.3
- CharXiv (RQ)tie77.577.5tie
MathKimi K2.5
- HMMT Feb. 2026Kimi K2.5 by 5.387.181.8+5.3
- IMOAnswerBenchKimi K2.5 by 581.876.8+5
- AIME 2026Kimi K2.5 by 2.595.893.3+2.5
Long documentsKimi K2.5
- AA-LCRKimi K2.5 by 67872+6
- LongBench v2Kimi K2.5 by 26159+2
Following instructionsEven
- IFBenchQwen3.5 35B A3B by 2.370.272.5+2.3
- Multi-ChallengeKimi K2.5 by 1.461.460+1.4
Other results28 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-OmniscienceKimi K2.5 by 40.8-7.3-48.1+40.8
- WideSearchKimi K2.5 by 21.97957.1+21.9
- Seal-0Kimi K2.5 by 1657.441.4+16
- SimpleVQAKimi K2.5 by 12.971.258.3+12.9
- NL2RepoKimi K2.5 by 11.53220.5+11.5
- Terminal-Bench 2.0Kimi K2.5 by 10.350.840.5+10.3
- AA Agentic IndexKimi K2.5 by 9.921.711.8+9.9
- Artificial Analysis Coding IndexKimi K2.5 by 9.846.837+9.8
- MMVUKimi K2.5 by 8.180.472.3+8.1
- vectara_answer_rateQwen3.5 35B A3B by 7.692.299.8+7.6
- BrowseComp-ZHQwen3.5 35B A3B by 7.262.369.5+7.2
- τ²-Bench Telecom (AA run)Kimi K2.5 by 6.795.989.2+6.7
- HMMT Feb. 2025Kimi K2.5 by 6.495.489+6.4
- HMMT 2025Kimi K2.5 by 6.295.489.2+6.2
- VideoMMMUKimi K2.5 by 6.286.680.4+6.2
- LVBenchKimi K2.5 by 4.575.971.4+4.5
- AA IntelligenceKimi K2.5 by 4.223.519.3+4.2
- vectara_factual_consistencyQwen3.5 35B A3B by 3.785.889.5+3.7
- ZeroBenchKimi K2.5 by 3118+3
- τ³-BenchQwen3.5 35B A3B by 2.96668.9+2.9
- MCPMarkKimi K2.5 by 2.529.527+2.5
- BabyVisionQwen3.5 35B A3B by 1.936.538.4+1.9
- MMLU-ProKimi K2.5 by 1.887.185.3+1.8
- vectara_avg_summary_lengthKimi K2.5 by 17.1 rating points11294.9+17.1 rating
- Tool-Decathlontie27.828.7tie
- OmniDocBench 1.5tie88.889.3tie
- MATH-Visiontie84.283.9tie
- HMMT Nov. 2025tie89.289.2tie
Questions people ask
Which is better, Kimi K2.5 or Qwen3.5 35B A3B?
Kimi K2.5 wins seven of the eight areas we test: coding, agents, reasoning, facts, images and charts, math and long documents. Qwen3.5 35B A3B wins none, but costs 38% less. They are level on following instructions.
Which is better for coding?
Kimi K2.5. It wins 8 of the 8 coding tests both models report; Qwen3.5 35B A3B wins none.
Which is cheaper?
Kimi K2.5 costs $0.60 per million input tokens and $3.00 per million output tokens; Qwen3.5 35B A3B costs $0.25 and $2.00. That makes Qwen3.5 35B A3B about 38% cheaper for the same work.
How do you compare the two?
We use the 58 benchmark tests both models have published scores on. The verdict counts the 30 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 28 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.