Qwen3.5 397B A17B vs Qwen3.6 27B
Wins 3 of 8 areas
Reasoning · Images and charts · Following instructions
Wins 3 of 8 areas
Coding · Agents · Math
The two are evenly matched.Qwen3.5 397B A17B is better at reasoning and images and charts; Qwen3.6 27B at agents and coding.
Scores updated · 41 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.
- ReasoningHard problems that need careful thinking20Qwen3.5 397B A17B2 of 3 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video20Qwen3.5 397B A17B2 of 2 tests
- Following instructionsDoing exactly what it is asked10Qwen3.5 397B A17B1 of 1 test
- AgentsCarrying out multi-step tasks on its own02Qwen3.6 27B2 of 3 tests · 1 tie
- CodingWriting and fixing software23Qwen3.6 27B3 of 7 tests · 2 ties
- MathCompetition and research-level math12Qwen3.6 27B2 of 4 tests · 1 tie
- FactsGetting facts right instead of making them up11Even1 each
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
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.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Qwen3.5 397B A17B pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+11.2points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+11.2points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+6.1points ahead
Where Qwen3.6 27B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+39.6points ahead
- Real work tasks from 44 professionsGDPVal+9.6points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+9.4points ahead
Every test, side by side
All 41 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 9.451.360.7+9.4
- Terminal-Bench HardQwen3.5 397B A17B by 6.140.934.8+6.1
- SWE-bench ProQwen3.6 27B by 2.650.953.5+2.6
- SciCodeQwen3.5 397B A17B by 244.842.8+2
- SWE-bench MultilingualQwen3.6 27B by 269.371.3+2
- SWE-bench Verifiedtie76.477.2tie
- LiveCodeBench v6tie83.683.9tie
AgentsQwen3.6 27B
- GDPValQwen3.6 27B by 9.614.824.4+9.6
- τ-Bench V3 · BankingQwen3.6 27B by 3.313.416.7+3.3
- Terminal-Bench 4.0tie00tie
ReasoningQwen3.5 397B A17B
- Humanity's Last ExamQwen3.5 397B A17B by 5.92923.1+5.9
- GPQA DiamondQwen3.5 397B A17B by 5.189.384.2+5.1
- CritPttie1.71.1tie
FactsEven
- AA-Omniscience · Non-hallucinationQwen3.6 27B by 39.611.150.7+39.6
- AA-Omniscience · AccuracyQwen3.5 397B A17B by 11.230.819.6+11.2
Images and chartsQwen3.5 397B A17B
MathQwen3.6 27B
- FrontierMath Tiers 1-3 (v2)Qwen3.6 27B by 3.931.235.1+3.9
- HMMT Feb. 2026Qwen3.5 397B A17B by 3.687.984.3+3.6
- AIME 2026Qwen3.6 27B by 2.891.394.1+2.8
- IMOAnswerBenchtie80.980.8tie
Following instructionsQwen3.5 397B A17B
- IFBenchQwen3.5 397B A17B by 11.278.867.6+11.2
Other results18 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 301 rating points11861487+301 rating
- SkillsBench Avg5Qwen3.6 27B by 18.23048.2+18.2
- AA-OmniscienceQwen3.6 27B by 10.7-30.8-20+10.7
- AA Agentic IndexQwen3.6 27B by 9.510.620.1+9.5
- Terminal-Bench 2.0Qwen3.6 27B by 6.852.559.3+6.8
- Artificial Analysis Coding IndexQwen3.6 27B by 5.548.253.7+5.5
- SuperGPQAQwen3.5 397B A17B by 4.470.466+4.4
- NL2RepoQwen3.6 27B by 432.236.2+4
- AA IntelligenceQwen3.6 27B by 318.421.4+3
- HMMT 2025Qwen3.5 397B A17B by 292.790.7+2
- HMMT Nov. 2025Qwen3.5 397B A17B by 292.790.7+2
- Claw-Eval AvgQwen3.6 27B by 1.770.772.4+1.7
- C-EvalQwen3.5 397B A17B by 1.69391.4+1.6
- MMLU-ProQwen3.5 397B A17B by 1.687.886.2+1.6
- QwenClawBenchQwen3.6 27B by 1.651.853.4+1.6
- MMLU-ReduxQwen3.5 397B A17B by 1.494.993.5+1.4
- τ²-Bench Telecom (AA run)Qwen3.5 397B A17B by 1.495.694.2+1.4
- HMMT Feb. 2025Qwen3.5 397B A17B by 194.893.8+1
Questions people ask
Which is better, Qwen3.5 397B A17B or Qwen3.6 27B?
Qwen3.5 397B A17B and Qwen3.6 27B each win three of the eight areas we test. Qwen3.5 397B A17B wins reasoning, images and charts and following instructions; Qwen3.6 27B wins coding, agents and math. They are level on facts and long documents.
Which is better for coding?
Qwen3.6 27B. It wins 3 of the 7 coding tests both models report; Qwen3.5 397B A17B wins 2, and 2 are ties.
Which is cheaper?
Qwen3.5 397B A17B costs $0.60 per million input tokens and $3.60 per million output tokens; Qwen3.6 27B costs $0.60 and $3.60. For a million tokens read plus a million written, they cost about the same.
How do you compare the two?
We use the 41 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 18 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.