Qwen3.6 27B vs Qwen3.6 35B A3B
Wins 7 of 8 areas
Coding · Agents · Facts · Images and charts · Math · Long documents · Following instructions
Wins 0 of 8 areas
—
Qwen3.6 27B is the stronger all-rounder.Qwen3.6 35B A3B is cheaper.
Scores updated · 54 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 software60Qwen3.6 27B6 of 7 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video30Qwen3.6 27B3 of 5 tests · 2 ties
- MathCompetition and research-level math30Qwen3.6 27B3 of 4 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own20Qwen3.6 27B2 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up10Qwen3.6 27B1 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts10Qwen3.6 27B1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3.6 27B1 of 1 test
- ReasoningHard problems that need careful thinking00Even0 each · 3 ties
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 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 Qwen3.6 27B pulls ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+15.8points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+14.7points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+7.4points ahead
Where Qwen3.6 35B A3B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 54 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3.6 27B
- Terminal-Bench 2.1Qwen3.6 27B by 15.860.744.9+15.8
- SciCodeQwen3.6 27B by 6.242.836.6+6.2
- SWE-bench MultilingualQwen3.6 27B by 4.171.367.2+4.1
- SWE-bench ProQwen3.6 27B by 453.549.5+4
- SWE-bench VerifiedQwen3.6 27B by 3.877.273.4+3.8
- LiveCodeBench v6Qwen3.6 27B by 3.583.980.4+3.5
- Terminal-Bench Hardtie34.834.8tie
AgentsQwen3.6 27B
- τ-Bench V3 · BankingQwen3.6 27B by 7.416.79.3+7.4
- GDPValQwen3.6 27B by 4.624.419.8+4.6
- Terminal-Bench 4.0tie00tie
ReasoningEven
- Humanity's Last Examtie23.122.2tie
- CritPttie1.10.3tie
- GPQA Diamondtie84.284.1tie
FactsQwen3.6 27B
- AA-Omniscience · Non-hallucinationQwen3.6 27B by 1.250.749.5+1.2
- AA-Omniscience · Accuracytie19.618.8tie
Images and chartsQwen3.6 27B
- OCRBenchv2Qwen3.6 27B by 2.868.365.5+2.8
- MMMUQwen3.6 27B by 1.282.981.7+1.2
- MathVistaQwen3.6 27B by 187.486.4+1
- CharXiv (RQ)tie78.478tie
- MMMU-Protie74.675tie
MathQwen3.6 27B
- FrontierMath Tiers 1-3 (v2)Qwen3.6 27B by 14.735.120.4+14.7
- IMOAnswerBenchQwen3.6 27B by 1.980.878.9+1.9
- AIME 2026Qwen3.6 27B by 1.494.192.7+1.4
- HMMT Feb. 2026tie84.383.6tie
Following instructionsQwen3.6 27B
- IFBenchQwen3.6 27B by 3.267.664.4+3.2
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.
- SkillsBench Avg5Qwen3.6 27B by 19.548.228.7+19.5
- Artificial Analysis Coding IndexQwen3.6 27B by 11.853.741.9+11.8
- Claw Eval (pass@3)Qwen3.6 27B by 10.660.650+10.6
- QwenWebBenchQwen3.6 27B by 90 rating points14871397+90 rating
- Terminal-Bench 2.0Qwen3.6 27B by 7.859.351.5+7.8
- NL2RepoQwen3.6 27B by 6.836.229.4+6.8
- RefSpatial-BenchQwen3.6 27B by 5.77064.3+5.7
- AA Agentic IndexQwen3.6 27B by 5.120.115+5.1
- Claw-Eval AvgQwen3.6 27B by 3.772.468.7+3.7
- AA IntelligenceQwen3.6 27B by 3.221.418.2+3.2
- HMMT Feb. 2025Qwen3.6 27B by 3.193.890.7+3.1
- SimpleVQAQwen3.6 35B A3B by 2.856.158.9+2.8
- AA-OmniscienceQwen3.6 27B by 2.2-20-22.2+2.2
- OCRBench v2 (Chinese)Qwen3.6 35B A3B by 2.158.760.8+2.1
- HMMT 2025Qwen3.6 27B by 1.690.789.1+1.6
- HMMT Nov. 2025Qwen3.6 27B by 1.690.789.1+1.6
- C-EvalQwen3.6 27B by 1.491.490+1.4
- SuperGPQAQwen3.6 27B by 1.36664.7+1.3
- RealWorldQAQwen3.6 35B A3B by 1.284.185.3+1.2
- τ²-Bench Telecom (AA run)Qwen3.6 35B A3B by 1.194.295.3+1.1
- MMLU-ProQwen3.6 27B by 186.285.2+1
- MV-Benchtie75.574.6tie
- QwenClawBenchtie53.452.6tie
- CC-OCRtie81.281.9tie
- VideoMMMUtie84.483.7tie
- MLVUtie86.686.2tie
- EmbSpatial-Benchtie84.684.3tie
- MMLU-Reduxtie93.593.3tie
Questions people ask
Which is better, Qwen3.6 27B or Qwen3.6 35B A3B?
Qwen3.6 27B wins seven of the eight areas we test: coding, agents, facts, images and charts, math, long documents and following instructions. Qwen3.6 35B A3B wins none, but costs 38% less. They are level on reasoning.
Which is better for coding?
Qwen3.6 27B. It wins 6 of the 7 coding tests both models report; Qwen3.6 35B A3B wins none, and 1 is a tie.
Which is cheaper?
Qwen3.6 27B costs $0.60 per million input tokens and $3.60 per million output tokens; Qwen3.6 35B A3B costs $0.38 and $2.25. That makes Qwen3.6 35B A3B about 38% cheaper for the same work.
How do you compare the two?
We use the 54 benchmark tests both models have published scores on. The verdict counts the 26 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.