Kimi K2.5 vs Qwen3.5 27B
Wins 5 of 8 areas
Coding · Reasoning · Facts · Images and charts · Math
Wins 1 of 8 areas
Following instructions
Kimi K2.5 is the stronger all-rounder.Qwen3.5 27B is cheaper and better at following instructions.
Both rank among the ten best models we track in images and charts.
Scores updated · 57 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 software60Kimi K2.56 of 7 tests · 1 tie
- ReasoningHard problems that need careful thinking30Kimi K2.53 of 3 tests
- MathCompetition and research-level math30Kimi K2.53 of 3 tests
- Images and chartsUnderstanding pictures, charts and video31Kimi K2.53 of 5 tests · 1 tie
- FactsGetting facts right instead of making them up21Kimi K2.52 of 3 tests
- Following instructionsDoing exactly what it is asked01Qwen3.5 27B1 of 2 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own22Even2 each
- Long documentsFinding answers in very long texts00Even0 each · 2 ties
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 27B costs 25% 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
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+15.8points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+14.5points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+13.9points ahead
Every test, side by side
All 57 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.5
- SciCodeKimi K2.5 by 9.54939.5+9.5
- LMArena · WebDevKimi K2.5 by 79 rating points14361357+79 rating
- SWE-bench VerifiedKimi K2.5 by 4.476.872.4+4.4
- LiveCodeBench v6Kimi K2.5 by 4.38580.7+4.3
- SWE-bench MultilingualKimi K2.5 by 3.77369.3+3.7
- Terminal-Bench HardKimi K2.5 by 2.234.832.6+2.2
- SWE-bench Protie50.751.2tie
AgentsEven
- GDPValQwen3.5 27B by 15.817.233+15.8
- BrowseCompKimi K2.5 by 13.974.961+13.9
- OSWorld-VerifiedKimi K2.5 by 7.163.356.2+7.1
- MCP AtlasQwen3.5 27B by 4.663.868.4+4.6
ReasoningKimi K2.5
- Humanity's Last ExamKimi K2.5 by 6.830.723.9+6.8
- CritPtKimi K2.5 by 2.23.10.9+2.2
- GPQA DiamondKimi K2.5 by 2.187.985.8+2.1
FactsKimi K2.5
- AA-Omniscience · Non-hallucinationKimi K2.5 by 15.834.318.5+15.8
- AA-Omniscience · AccuracyKimi K2.5 by 14.535.220.7+14.5
- Vectara HHEM hallucination ratelower is betterQwen3.5 27B by 2.114.212.1+2.1
Images and chartsKimi K2.5
- LMArena · VisionKimi K2.5 by 29 rating points12691240+29 rating
- OCRBenchKimi K2.5 by 2.992.389.4+2.9
- MathVistaKimi K2.5 by 2.390.187.8+2.3
- CharXiv (RQ)Qwen3.5 27B by 277.579.5+2
- MMMU-Protie75.475tie
MathKimi K2.5
- HMMT Feb. 2026Kimi K2.5 by 687.181.1+6
- AIME 2026Kimi K2.5 by 4.195.891.7+4.1
- IMOAnswerBenchKimi K2.5 by 1.981.879.9+1.9
Following instructionsQwen3.5 27B
- IFBenchQwen3.5 27B by 5.470.275.6+5.4
- Multi-Challengetie61.460.8tie
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 36.7-7.3-44+36.7
- AA Agentic IndexQwen3.5 27B by 32.921.754.6+32.9
- WideSearchKimi K2.5 by 17.97961.1+17.9
- SimpleVQAKimi K2.5 by 15.271.256+15.2
- Artificial Analysis Coding IndexKimi K2.5 by 11.946.834.9+11.9
- Seal-0Kimi K2.5 by 10.257.447.2+10.2
- Terminal-Bench 2.0Kimi K2.5 by 9.250.841.6+9.2
- BabyVisionQwen3.5 27B by 8.136.544.6+8.1
- vectara_answer_rateQwen3.5 27B by 7.692.299.8+7.6
- MMVUKimi K2.5 by 7.180.473.3+7.1
- MCPMarkQwen3.5 27B by 6.829.536.3+6.8
- HMMT 2025Kimi K2.5 by 5.695.489.8+5.6
- NL2RepoKimi K2.5 by 4.73227.3+4.7
- VideoMMMUKimi K2.5 by 4.386.682.3+4.3
- Tool-DecathlonQwen3.5 27B by 3.727.831.5+3.7
- HMMT Feb. 2025Kimi K2.5 by 3.495.492+3.4
- τ³-BenchQwen3.5 27B by 2.46668.4+2.4
- LVBenchKimi K2.5 by 2.375.973.6+2.3
- vectara_factual_consistencyQwen3.5 27B by 2.185.887.9+2.1
- τ²-Bench Telecom (AA run)Kimi K2.5 by 295.993.9+2
- MATH-VisionQwen3.5 27B by 1.884.286+1.8
- vectara_avg_summary_lengthKimi K2.5 by 17.6 rating points11294.4+17.6 rating
- MMLU-ProKimi K2.5 by 187.186.1+1
- ZeroBenchKimi K2.5 by 11110+1
- AA Intelligencetie23.522.9tie
- HMMT Nov. 2025tie89.289.8tie
- BrowseComp-ZHtie62.362.1tie
- OmniDocBench 1.5tie88.888.9tie
Questions people ask
Which is better, Kimi K2.5 or Qwen3.5 27B?
Kimi K2.5 wins five of the eight areas we test: coding, reasoning, facts, images and charts and math. Qwen3.5 27B wins following instructions, and costs 25% less. They are level on agents and long documents.
Which is better for coding?
Kimi K2.5. It wins 6 of the 7 coding tests both models report; Qwen3.5 27B wins none, and 1 is a tie.
Which is cheaper?
Kimi K2.5 costs $0.60 per million input tokens and $3.00 per million output tokens; Qwen3.5 27B costs $0.30 and $2.40. That makes Qwen3.5 27B about 25% cheaper for the same work.
How do you compare the two?
We use the 57 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 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.