Qwen3.5 27B vs Qwen3.6 35B A3B
Wins 6 of 8 areas
Coding · Agents · Reasoning · Images and charts · Long documents · Following instructions
Wins 1 of 8 areas
Math
Qwen3.5 27B is the stronger all-rounder.Qwen3.6 35B A3B is cheaper and better at math.
Scores updated · 61 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.
- AgentsCarrying out multi-step tasks on its own30Qwen3.5 27B3 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20Qwen3.5 27B2 of 4 tests · 2 ties
- ReasoningHard problems that need careful thinking20Qwen3.5 27B2 of 3 tests · 1 tie
- CodingWriting and fixing software32Qwen3.5 27B3 of 6 tests · 1 tie
- Long documentsFinding answers in very long texts10Qwen3.5 27B1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3.5 27B1 of 1 test
- MathCompetition and research-level math12Qwen3.6 35B A3B2 of 3 tests
- FactsGetting facts right instead of making them up11Even1 each
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.
Qwen3.6 35B A3B costs 3% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Qwen3.5 27B pulls ahead
- Real work tasks from 44 professionsGDPVal+13.2points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+12.3points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+11.2points ahead
Where Qwen3.6 35B A3B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+31points ahead
- Harvard-MIT high-school math contest problemsHMMT Feb. 2026+2.5points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+2.2points ahead
Every test, side by side
All 61 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3.5 27B
- SciCodeQwen3.5 27B by 2.939.536.6+2.9
- Terminal-Bench HardQwen3.6 35B A3B by 2.232.634.8+2.2
- SWE-bench MultilingualQwen3.5 27B by 2.169.367.2+2.1
- SWE-bench ProQwen3.5 27B by 1.751.249.5+1.7
- SWE-bench VerifiedQwen3.6 35B A3B by 172.473.4+1
- LiveCodeBench v6tie80.780.4tie
AgentsQwen3.5 27B
- GDPValQwen3.5 27B by 13.23319.8+13.2
- BrowseCompQwen3.5 27B by 12.36148.7+12.3
- MCP AtlasQwen3.5 27B by 5.668.462.8+5.6
ReasoningQwen3.5 27B
- GPQA DiamondQwen3.5 27B by 1.785.884.1+1.7
- Humanity's Last ExamQwen3.5 27B by 1.723.922.2+1.7
- CritPttie0.90.3tie
FactsEven
- AA-Omniscience · Non-hallucinationQwen3.6 35B A3B by 3118.549.5+31
- AA-Omniscience · AccuracyQwen3.5 27B by 1.920.718.8+1.9
Images and chartsQwen3.5 27B
- CharXiv (RQ)Qwen3.5 27B by 1.579.578+1.5
- MathVistaQwen3.5 27B by 1.487.886.4+1.4
- MMMUtie82.381.7tie
- MMMU-Protie7575tie
MathQwen3.6 35B A3B
- HMMT Feb. 2026Qwen3.6 35B A3B by 2.581.183.6+2.5
- AIME 2026Qwen3.6 35B A3B by 191.792.7+1
- IMOAnswerBenchQwen3.5 27B by 179.978.9+1
Following instructionsQwen3.5 27B
- IFBenchQwen3.5 27B by 11.275.664.4+11.2
Other results38 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 Agentic IndexQwen3.5 27B by 39.654.615+39.6
- QwenWebBenchQwen3.6 35B A3B by 329 rating points10681397+329 rating
- AA-OmniscienceQwen3.6 35B A3B by 21.8-44-22.2+21.8
- Terminal-Bench 2.0Qwen3.6 35B A3B by 9.941.651.5+9.9
- Artificial Analysis Coding IndexQwen3.6 35B A3B by 734.941.9+7
- VITA-BenchQwen3.5 27B by 6.341.935.6+6.3
- AA IntelligenceQwen3.5 27B by 4.722.918.2+4.7
- Tool-DecathlonQwen3.5 27B by 4.631.526.9+4.6
- Claw-Eval AvgQwen3.6 35B A3B by 4.464.368.7+4.4
- Claw Eval (pass@3)Qwen3.6 35B A3B by 3.846.250+3.8
- RefSpatial-BenchQwen3.5 27B by 3.467.764.3+3.4
- DeepPlanningQwen3.6 35B A3B by 3.322.625.9+3.3
- SimpleVQAQwen3.6 35B A3B by 2.95658.9+2.9
- LVBenchQwen3.5 27B by 2.273.671.4+2.2
- NL2RepoQwen3.6 35B A3B by 2.127.329.4+2.1
- RealWorldQAQwen3.6 35B A3B by 1.683.785.3+1.6
- SkillsBench Avg5Qwen3.6 35B A3B by 1.527.228.7+1.5
- VideoMMMUQwen3.6 35B A3B by 1.482.383.7+1.4
- τ²-Bench Telecom (AA run)Qwen3.6 35B A3B by 1.493.995.3+1.4
- HMMT Feb. 2025Qwen3.5 27B by 1.39290.7+1.3
- τ³-BenchQwen3.5 27B by 1.268.467.2+1.2
- OmniDocBench 1.5Qwen3.6 35B A3B by 188.989.9+1
- WideSearchQwen3.5 27B by 161.160.1+1
- CC-OCRtie8181.9tie
- MMLU-Protie86.185.2tie
- SuperGPQAtie65.664.7tie
- HMMT 2025tie89.889.1tie
- HMMT Nov. 2025tie89.889.1tie
- MCPMarktie36.337tie
- C-Evaltie90.590tie
- QwenClawBenchtie52.252.6tie
- MLVUtie85.986.2tie
- AI2Dtie92.992.7tie
- EmbSpatial-Benchtie84.584.3tie
- HallusionBenchtie7069.8tie
- MMBench EN-DEV-v1.1tie92.692.8tie
- MMLU-Reduxtie93.293.3tie
- MV-Benchtie74.674.6tie
Questions people ask
Which is better, Qwen3.5 27B or Qwen3.6 35B A3B?
Qwen3.5 27B wins six of the eight areas we test: coding, agents, reasoning, images and charts, long documents and following instructions. Qwen3.6 35B A3B wins math, and costs 3% less. They are level on facts.
Which is better for coding?
Qwen3.5 27B. It wins 3 of the 6 coding tests both models report; Qwen3.6 35B A3B wins 2, and 1 is a tie.
Which is cheaper?
Qwen3.5 27B costs $0.30 per million input tokens and $2.40 per million output tokens; Qwen3.6 35B A3B costs $0.38 and $2.25. That makes Qwen3.6 35B A3B about 3% cheaper for the same work.
How do you compare the two?
We use the 61 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 38 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.