Kimi K2.5 vs Qwen3.5 122B A10B
Wins 6 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Long documents
Wins 1 of 7 areas
Following instructions
Kimi K2.5 is the stronger all-rounder.Qwen3.5 122B A10B is better at following instructions.
Both rank among the ten best models we track in images and charts.
Scores updated · 51 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 software71Kimi K2.57 of 8 tests
- ReasoningHard problems that need careful thinking30Kimi K2.53 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20Kimi K2.52 of 5 tests · 3 ties
- AgentsCarrying out multi-step tasks on its own31Kimi K2.53 of 4 tests
- FactsGetting facts right instead of making them up21Kimi K2.52 of 3 tests
- Long documentsFinding answers in very long texts10Kimi K2.51 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked01Qwen3.5 122B A10B1 of 2 tests · 1 tie
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.
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+21.4points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+11.1points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+11points ahead
Where Qwen3.5 122B A10B pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+5.5points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+1.9points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+1.1points ahead
Every test, side by side
All 51 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.5
- SWE-bench MultilingualKimi K2.5 by 117362+11
- SciCodeKimi K2.5 by 9.34939.7+9.3
- LMArena · WebDevKimi K2.5 by 77 rating points14371360+77 rating
- LiveCodeBench v6Kimi K2.5 by 6.18578.9+6.1
- SWE-bench ProKimi K2.5 by 5.850.744.9+5.8
- SWE-bench VerifiedKimi K2.5 by 4.876.872+4.8
- Terminal-Bench HardKimi K2.5 by 3.734.831.1+3.7
- Terminal-Bench 2.1Qwen3.5 122B A10B by 1.945.747.6+1.9
AgentsKimi K2.5
- BrowseCompKimi K2.5 by 11.174.963.8+11.1
- OSWorld-VerifiedKimi K2.5 by 5.363.358+5.3
- GDPValKimi K2.5 by 1.31715.7+1.3
- τ-Bench V3 · BankingQwen3.5 122B A10B by 1.114.215.3+1.1
ReasoningKimi K2.5
- Humanity's Last ExamKimi K2.5 by 5.530.725.2+5.5
- CritPtKimi K2.5 by 2.53.10.6+2.5
- GPQA DiamondKimi K2.5 by 2.287.985.7+2.2
FactsKimi K2.5
- AA-Omniscience · Non-hallucinationKimi K2.5 by 21.434.312.9+21.4
- AA-Omniscience · AccuracyKimi K2.5 by 10.835.224.4+10.8
- Vectara HHEM hallucination ratelower is betterQwen3.5 122B A10B by 314.211.2+3
Images and chartsKimi K2.5
- MathVistaKimi K2.5 by 2.790.187.4+2.7
- LMArena · VisionKimi K2.5 by 24 rating points12691245+24 rating
- MMMU-Protie75.475tie
- CharXiv (RQ)tie77.577.2tie
- OCRBenchtie92.392.1tie
Long documentsKimi K2.5
- AA-LCRKimi K2.5 by 1.77876.3+1.7
- LongBench v2tie6160.2tie
Following instructionsQwen3.5 122B A10B
- IFBenchQwen3.5 122B A10B by 5.570.275.7+5.5
- Multi-Challengetie61.461.5tie
Other results24 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 34.2-7.3-41.5+34.2
- WideSearchKimi K2.5 by 18.57960.5+18.5
- GDPval-AA v2Kimi K2.5 by 148 rating points1009861+148 rating
- Seal-0Kimi K2.5 by 13.357.444.1+13.3
- AA Agentic IndexKimi K2.5 by 12.121.79.6+12.1
- SimpleVQAKimi K2.5 by 9.571.261.7+9.5
- AA IntelligenceKimi K2.5 by 7.923.515.6+7.9
- BrowseComp-ZHQwen3.5 122B A10B by 7.662.369.9+7.6
- vectara_answer_rateQwen3.5 122B A10B by 7.692.299.8+7.6
- MMVUKimi K2.5 by 5.780.474.7+5.7
- HMMT 2025Kimi K2.5 by 5.195.490.3+5.1
- VideoMMMUKimi K2.5 by 4.686.682+4.6
- BabyVisionQwen3.5 122B A10B by 3.736.540.2+3.7
- vectara_factual_consistencyQwen3.5 122B A10B by 385.888.8+3
- vectara_avg_summary_lengthKimi K2.5 by 25.6 rating points11286.4+25.6 rating
- τ²-Bench Telecom (AA run)Kimi K2.5 by 2.395.993.6+2.3
- MATH-VisionQwen3.5 122B A10B by 284.286.2+2
- ZeroBenchKimi K2.5 by 2119+2
- LVBenchKimi K2.5 by 1.575.974.4+1.5
- Terminal-Bench 2.0Kimi K2.5 by 1.450.849.4+1.4
- Artificial Analysis Coding IndexKimi K2.5 by 1.146.845.7+1.1
- OmniDocBench 1.5Qwen3.5 122B A10B by 188.889.8+1
- MMLU-Protie87.186.7tie
- τ³-Bench Bankingtie14.214.2tie
Questions people ask
Which is better, Kimi K2.5 or Qwen3.5 122B A10B?
Kimi K2.5 wins six of the seven areas where both have results: coding, agents, reasoning, facts, images and charts and long documents. Qwen3.5 122B A10B wins following instructions.
Which is better for coding?
Kimi K2.5. It wins 7 of the 8 coding tests both models report; Qwen3.5 122B A10B wins 1.
Which is cheaper?
Kimi K2.5 costs $0.60 per million input tokens and $3.00 per million output tokens; Qwen3.5 122B A10B costs $0.40 and $3.20. For a million tokens read plus a million written, they cost about the same.
How do you compare the two?
We use the 51 benchmark tests both models have published scores on. The verdict counts the 27 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 24 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.