Kimi K2.6 vs Qwen3.5 35B A3B
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
Wins 0 of 8 areas
—
Kimi K2.6 is the stronger all-rounder.Qwen3.5 35B A3B is cheaper.
Scores updated · 52 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 K2.68 of 8 tests
- AgentsCarrying out multi-step tasks on its own60Kimi K2.66 of 6 tests
- ReasoningHard problems that need careful thinking30Kimi K2.63 of 3 tests
- MathCompetition and research-level math30Kimi K2.63 of 3 tests
- FactsGetting facts right instead of making them up20Kimi K2.62 of 3 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video20Kimi K2.62 of 2 tests
- Long documentsFinding answers in very long texts10Kimi K2.61 of 1 test
- Following instructionsDoing exactly what it is asked10Kimi K2.61 of 1 test
Long documents and following instructions rest on a single test 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 55% 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 K2.6 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+44.9points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+25.3points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+25.1points ahead
Where Qwen3.5 35B A3B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 52 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.6
- LMArena · WebDevKimi K2.6 by 259 rating points15091250+259 rating
- Terminal-Bench 2.1Kimi K2.6 by 25.165.940.8+25.1
- Terminal-Bench HardKimi K2.6 by 17.443.926.5+17.4
- SWE-bench MultilingualKimi K2.6 by 16.476.760.3+16.4
- LiveCodeBench v6Kimi K2.6 by 1589.674.6+15
- SWE-bench ProKimi K2.6 by 1458.644.6+14
- SciCodeKimi K2.6 by 13.851.537.7+13.8
- SWE-bench VerifiedKimi K2.6 by 1180.269.2+11
AgentsKimi K2.6
- BrowseCompKimi K2.6 by 25.386.361+25.3
- GDPValKimi K2.6 by 21.4275.6+21.4
- OSWorld-VerifiedKimi K2.6 by 18.673.154.5+18.6
- τ-Bench V3 · BankingKimi K2.6 by 18.423.34.9+18.4
- AA IT-Bench SREKimi K2.6 by 9.731.221.5+9.7
- MCP AtlasKimi K2.6 by 5.768.162.4+5.7
ReasoningKimi K2.6
- Humanity's Last ExamKimi K2.6 by 16.537.521+16.5
- CritPtKimi K2.6 by 7.180.9+7.1
- GPQA DiamondKimi K2.6 by 6.691.184.5+6.6
FactsKimi K2.6
- AA-Omniscience · Non-hallucinationKimi K2.6 by 44.959.514.6+44.9
- AA-Omniscience · AccuracyKimi K2.6 by 12.532.620.1+12.5
- Vectara HHEM hallucination ratelower is bettertie10.810.5tie
Images and chartsKimi K2.6
- CharXiv (RQ)Kimi K2.6 by 9.286.777.5+9.2
- MMMU-ProKimi K2.6 by 6.779.472.7+6.7
MathKimi K2.6
- HMMT Feb. 2026Kimi K2.6 by 10.992.781.8+10.9
- IMOAnswerBenchKimi K2.6 by 9.28676.8+9.2
- AIME 2026Kimi K2.6 by 3.196.493.3+3.1
Other results25 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 K2.6 by 53.45.3-48.1+53.4
- BabyVisionKimi K2.6 by 30.168.538.4+30.1
- Claw Eval (pass@3)Kimi K2.6 by 29.980.951+29.9
- MCPMarkKimi K2.6 by 28.955.927+28.9
- Terminal-Bench 2.0Kimi K2.6 by 26.266.740.5+26.2
- Artificial Analysis Coding IndexKimi K2.6 by 24.861.837+24.8
- OJBenchKimi K2.6 by 24.660.636+24.6
- WideSearchKimi K2.6 by 23.780.857.1+23.7
- AgentWorldBench - SWEKimi K2.6 by 11.258.847.6+11.2
- AA Agentic IndexKimi K2.6 by 10.322.111.8+10.3
- MATH-VisionKimi K2.6 by 9.393.283.9+9.3
- AA IntelligenceKimi K2.6 by 7.72719.3+7.7
- AgentWorldBench - MCPKimi K2.6 by 7.365.257.9+7.3
- GPQA (unspecified)Kimi K2.6 by 6.89184.2+6.8
- τ²-Bench Telecom (AA run)Kimi K2.6 by 6.795.989.2+6.7
- AgentWorldBench - TerminalKimi K2.6 by 6.452.546.1+6.4
- AgentWorldBench - AndroidKimi K2.6 by 5.758.953.2+5.7
- AgentWorldBench - OverallKimi K2.6 by 5.753.447.7+5.7
- AgentWorldBench - OSKimi K2.6 by 4.560.856.3+4.5
- AgentWorldBench - WebKimi K2.6 by 3.150.247.1+3.1
- vectara_avg_summary_lengthKimi K2.6 by 21.8 rating points116.794.9+21.8 rating
- MMLU-ProKimi K2.6 by 1.887.185.3+1.8
- AgentWorldBench - SearchKimi K2.6 by 1.527.526+1.5
- vectara_factual_consistencytie89.289.5tie
- vectara_answer_ratetie99.799.8tie
Questions people ask
Which is better, Kimi K2.6 or Qwen3.5 35B A3B?
Kimi K2.6 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Qwen3.5 35B A3B wins none, but costs 55% less.
Which is better for coding?
Kimi K2.6. It wins 8 of the 8 coding tests both models report; Qwen3.5 35B A3B wins none.
Which is cheaper?
Kimi K2.6 costs $0.95 per million input tokens and $4.00 per million output tokens; Qwen3.5 35B A3B costs $0.25 and $2.00. That makes Qwen3.5 35B A3B about 55% cheaper for the same work.
How do you compare the two?
We use the 52 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 25 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.