DeepSeek-V3.2 vs Qwen3.5 35B A3B
Wins 6 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents
Wins 1 of 7 areas
Following instructions
DeepSeek-V3.2 is the stronger all-rounder.Qwen3.5 35B A3B is better at following instructions.
Scores updated · 45 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. 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 software71DeepSeek-V3.27 of 8 tests
- FactsGetting facts right instead of making them up30DeepSeek-V3.23 of 3 tests
- ReasoningHard problems that need careful thinking20DeepSeek-V3.22 of 3 tests · 1 tie
- MathCompetition and research-level math20DeepSeek-V3.22 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own21DeepSeek-V3.22 of 4 tests · 1 tie
- Long documentsFinding answers in very long texts10DeepSeek-V3.21 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked01Qwen3.5 35B A3B1 of 1 test
Following instructions rests on a single test.
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.
DeepSeek-V3.2 costs 69% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where DeepSeek-V3.2 pulls ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+13.9points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+12.9points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+9.9points ahead
Where Qwen3.5 35B A3B pulls ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+29points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+11.8points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+9.6points ahead
Every test, side by side
All 45 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3.2
- SWE-bench ProQwen3.5 35B A3B by 2915.644.6+29
- LMArena · WebDevDeepSeek-V3.2 by 112 rating points13621250+112 rating
- SWE-bench MultilingualDeepSeek-V3.2 by 9.970.260.3+9.9
- Terminal-Bench HardDeepSeek-V3.2 by 9.135.626.5+9.1
- LiveCodeBench v6DeepSeek-V3.2 by 8.783.374.6+8.7
- Terminal-Bench 2.1DeepSeek-V3.2 by 646.840.8+6
- SWE-bench VerifiedDeepSeek-V3.2 by 3.973.169.2+3.9
- SciCodeDeepSeek-V3.2 by 1.238.937.7+1.2
AgentsDeepSeek-V3.2
- τ-Bench V3 · BankingDeepSeek-V3.2 by 13.918.84.9+13.9
- BrowseCompQwen3.5 35B A3B by 9.651.461+9.6
- GDPValDeepSeek-V3.2 by 4.29.85.6+4.2
- MCP Atlastie62.262.4tie
ReasoningDeepSeek-V3.2
- Humanity's Last ExamDeepSeek-V3.2 by 3.624.621+3.6
- CritPtDeepSeek-V3.2 by 22.90.9+2
- GPQA Diamondtie8484.5tie
FactsDeepSeek-V3.2
- AA-Omniscience · AccuracyDeepSeek-V3.2 by 12.93320.1+12.9
- Vectara HHEM hallucination ratelower is betterDeepSeek-V3.2 by 4.26.310.5+4.2
- AA-Omniscience · Non-hallucinationDeepSeek-V3.2 by 2.717.314.6+2.7
MathDeepSeek-V3.2
- HMMT Feb. 2026DeepSeek-V3.2 by 2.384.181.8+2.3
- IMOAnswerBenchDeepSeek-V3.2 by 1.578.376.8+1.5
- AIME 2026tie94.293.3tie
Long documentsDeepSeek-V3.2
- AA-LCRDeepSeek-V3.2 by 1.373.372+1.3
- LongBench v2tie59.859tie
Following instructionsQwen3.5 35B A3B
- IFBenchQwen3.5 35B A3B by 11.860.772.5+11.8
Other results21 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.
- vectara_avg_summary_lengthQwen3.5 35B A3B by 32.96294.9+32.9
- AA-OmniscienceDeepSeek-V3.2 by 25.6-22.5-48.1+25.6
- MCPMarkDeepSeek-V3.2 by 113827+11
- Seal-0DeepSeek-V3.2 by 8.149.541.4+8.1
- Artificial Analysis Coding IndexDeepSeek-V3.2 by 7.244.237+7.2
- vectara_answer_rateQwen3.5 35B A3B by 7.292.699.8+7.2
- AA Agentic IndexDeepSeek-V3.2 by 6.518.311.8+6.5
- Tool-DecathlonDeepSeek-V3.2 by 6.535.228.7+6.5
- Terminal-Bench 2.0DeepSeek-V3.2 by 5.946.440.5+5.9
- BrowseComp-ZHQwen3.5 35B A3B by 4.56569.5+4.5
- GPQA (unspecified)Qwen3.5 35B A3B by 4.379.984.2+4.3
- vectara_factual_consistencyDeepSeek-V3.2 by 4.293.789.5+4.2
- HMMT Feb. 2025DeepSeek-V3.2 by 3.592.589+3.5
- AA IntelligenceDeepSeek-V3.2 by 2.221.519.3+2.2
- τ²-Bench Telecom (AA run)DeepSeek-V3.2 by 1.490.689.2+1.4
- HMMT 2025DeepSeek-V3.2 by 190.289.2+1
- t2-benchQwen3.5 35B A3B by 180.281.2+1
- HMMT Nov. 2025tie9089.2tie
- MMLU-Reduxtie93.793.3tie
- MMLU-Protie8585.3tie
- τ³-Benchtie69.268.9tie
Questions people ask
Which is better, DeepSeek-V3.2 or Qwen3.5 35B A3B?
DeepSeek-V3.2 wins six of the seven areas where both have results: coding, agents, reasoning, facts, math and long documents. Qwen3.5 35B A3B wins following instructions.
Which is better for coding?
DeepSeek-V3.2. It wins 7 of the 8 coding tests both models report; Qwen3.5 35B A3B wins 1.
Which is cheaper?
DeepSeek-V3.2 costs $0.28 per million input tokens and $0.42 per million output tokens; Qwen3.5 35B A3B costs $0.25 and $2.00. That makes DeepSeek-V3.2 about 69% cheaper for the same work.
How do you compare the two?
We use the 45 benchmark tests both models have published scores on. The verdict counts the 24 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 21 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.