Qwen3.5 27B vs Qwen3.6 27B
Wins 4 of 8 areas
Agents · Reasoning · Images and charts · Following instructions
Wins 2 of 8 areas
Coding · Math
Qwen3.5 27B wins more areas, narrowly.Qwen3.6 27B is better at coding.
Both rank among the ten best models we track in images and charts.
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.
- Images and chartsUnderstanding pictures, charts and video10Qwen3.5 27B1 of 6 tests · 5 ties
- ReasoningHard problems that need careful thinking10Qwen3.5 27B1 of 3 tests · 2 ties
- AgentsCarrying out multi-step tasks on its own10Qwen3.5 27B1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3.5 27B1 of 1 test
- CodingWriting and fixing software06Qwen3.6 27B6 of 6 tests
- MathCompetition and research-level math02Qwen3.6 27B2 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up11Even1 each
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
Agents, 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 27B costs 36% 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+8.6points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+8points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+1.6points ahead
Where Qwen3.6 27B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+32.2points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+4.8points ahead
- Code for real scientific research problemsSciCode+3.3points ahead
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.
CodingQwen3.6 27B
- SWE-bench VerifiedQwen3.6 27B by 4.872.477.2+4.8
- SciCodeQwen3.6 27B by 3.339.542.8+3.3
- LiveCodeBench v6Qwen3.6 27B by 3.280.783.9+3.2
- SWE-bench ProQwen3.6 27B by 2.351.253.5+2.3
- Terminal-Bench HardQwen3.6 27B by 2.232.634.8+2.2
- SWE-bench MultilingualQwen3.6 27B by 269.371.3+2
ReasoningQwen3.5 27B
- GPQA DiamondQwen3.5 27B by 1.685.884.2+1.6
- Humanity's Last Examtie23.923.1tie
- CritPttie0.91.1tie
FactsEven
- AA-Omniscience · Non-hallucinationQwen3.6 27B by 32.218.550.7+32.2
- AA-Omniscience · AccuracyQwen3.5 27B by 1.120.719.6+1.1
Images and chartsQwen3.5 27B
MathQwen3.6 27B
- HMMT Feb. 2026Qwen3.6 27B by 3.281.184.3+3.2
- AIME 2026Qwen3.6 27B by 2.491.794.1+2.4
- IMOAnswerBenchtie79.980.8tie
Following instructionsQwen3.5 27B
- IFBenchQwen3.5 27B by 875.667.6+8
Other results29 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.
- QwenWebBenchQwen3.6 27B by 419 rating points10681487+419 rating
- AA Agentic IndexQwen3.5 27B by 34.554.620.1+34.5
- AA-OmniscienceQwen3.6 27B by 24-44-20+24
- SkillsBench Avg5Qwen3.6 27B by 2127.248.2+21
- Artificial Analysis Coding IndexQwen3.6 27B by 18.834.953.7+18.8
- Terminal-Bench 2.0Qwen3.6 27B by 17.741.659.3+17.7
- Claw Eval (pass@3)Qwen3.6 27B by 14.446.260.6+14.4
- NL2RepoQwen3.6 27B by 8.927.336.2+8.9
- Claw-Eval AvgQwen3.6 27B by 8.164.372.4+8.1
- RefSpatial-BenchQwen3.6 27B by 2.367.770+2.3
- DynaMathQwen3.5 27B by 2.187.785.6+2.1
- VideoMMMUQwen3.6 27B by 2.182.384.4+2.1
- ERQAQwen3.6 27B by 260.562.5+2
- HMMT Feb. 2025Qwen3.6 27B by 1.89293.8+1.8
- AA IntelligenceQwen3.5 27B by 1.522.921.4+1.5
- QwenClawBenchQwen3.6 27B by 1.252.253.4+1.2
- C-Evaltie90.591.4tie
- HMMT 2025tie89.890.7tie
- HMMT Nov. 2025tie89.890.7tie
- MV-Benchtie74.675.5tie
- MLVUtie85.986.6tie
- RealWorldQAtie83.784.1tie
- SuperGPQAtie65.666tie
- MMLU-Reduxtie93.293.5tie
- τ²-Bench Telecom (AA run)tie93.994.2tie
- CC-OCRtie8181.2tie
- EmbSpatial-Benchtie84.584.6tie
- MMLU-Protie86.186.2tie
- SimpleVQAtie5656.1tie
Questions people ask
Which is better, Qwen3.5 27B or Qwen3.6 27B?
Qwen3.5 27B wins four of the eight areas we test: agents, reasoning, images and charts and following instructions. Qwen3.6 27B wins coding and math. They are level on facts and long documents.
Which is better for coding?
Qwen3.6 27B. It wins 6 of the 6 coding tests both models report; Qwen3.5 27B wins none.
Which is cheaper?
Qwen3.5 27B costs $0.30 per million input tokens and $2.40 per million output tokens; Qwen3.6 27B costs $0.60 and $3.60. That makes Qwen3.5 27B about 36% 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 23 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 29 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.