DeepSeek-V3.2 vs Qwen3.5 122B A10B
Wins 2 of 6 areas
Coding · Facts
Wins 3 of 6 areas
Agents · Long documents · Following instructions
Qwen3.5 122B A10B wins more areas, narrowly.DeepSeek-V3.2 is cheaper and better at coding.
Scores updated · 35 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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.
- AgentsCarrying out multi-step tasks on its own12Qwen3.5 122B A10B2 of 3 tests
- Long documentsFinding answers in very long texts01Qwen3.5 122B A10B1 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked01Qwen3.5 122B A10B1 of 1 test
- CodingWriting and fixing software41DeepSeek-V3.24 of 8 tests · 3 ties
- FactsGetting facts right instead of making them up30DeepSeek-V3.23 of 3 tests
- ReasoningHard problems that need careful thinking11Even1 each · 1 tie
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 81% 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 122B A10B pulls ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+29.3points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+15points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+12.4points ahead
Where DeepSeek-V3.2 pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+8.6points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+8.2points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+4.5points ahead
Every test, side by side
All 35 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 122B A10B by 29.315.644.9+29.3
- SWE-bench MultilingualDeepSeek-V3.2 by 8.270.262+8.2
- Terminal-Bench HardDeepSeek-V3.2 by 4.535.631.1+4.5
- LiveCodeBench v6DeepSeek-V3.2 by 4.483.378.9+4.4
- SWE-bench VerifiedDeepSeek-V3.2 by 1.173.172+1.1
- SciCodetie38.939.7tie
- Terminal-Bench 2.1tie46.847.6tie
- LMArena · WebDevtie13621360tie
AgentsQwen3.5 122B A10B
- BrowseCompQwen3.5 122B A10B by 12.451.463.8+12.4
- GDPValQwen3.5 122B A10B by 6.19.615.7+6.1
- τ-Bench V3 · BankingDeepSeek-V3.2 by 3.518.815.3+3.5
ReasoningEven
- CritPtDeepSeek-V3.2 by 2.32.90.6+2.3
- GPQA DiamondQwen3.5 122B A10B by 1.78485.7+1.7
- Humanity's Last Examtie24.625.2tie
FactsDeepSeek-V3.2
- AA-Omniscience · AccuracyDeepSeek-V3.2 by 8.63324.4+8.6
- Vectara HHEM hallucination ratelower is betterDeepSeek-V3.2 by 4.96.311.2+4.9
- AA-Omniscience · Non-hallucinationDeepSeek-V3.2 by 4.417.312.9+4.4
Long documentsQwen3.5 122B A10B
- AA-LCRQwen3.5 122B A10B by 373.376.3+3
- LongBench v2tie59.860.2tie
Following instructionsQwen3.5 122B A10B
- IFBenchQwen3.5 122B A10B by 1560.775.7+15
Other results15 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 122B A10B by 24.46286.4+24.4
- AA-OmniscienceDeepSeek-V3.2 by 19-22.5-41.5+19
- AA Agentic IndexDeepSeek-V3.2 by 8.718.39.6+8.7
- vectara_answer_rateQwen3.5 122B A10B by 7.292.699.8+7.2
- AA IntelligenceDeepSeek-V3.2 by 5.921.515.6+5.9
- Seal-0DeepSeek-V3.2 by 5.449.544.1+5.4
- BrowseComp-ZHQwen3.5 122B A10B by 4.96569.9+4.9
- vectara_factual_consistencyDeepSeek-V3.2 by 4.993.788.8+4.9
- Terminal-Bench 2.0Qwen3.5 122B A10B by 346.449.4+3
- τ²-Bench Telecom (AA run)Qwen3.5 122B A10B by 390.693.6+3
- MMLU-ProQwen3.5 122B A10B by 1.78586.7+1.7
- Artificial Analysis Coding IndexQwen3.5 122B A10B by 1.544.245.7+1.5
- t2-benchtie80.279.5tie
- MMLU-Reduxtie93.794tie
- HMMT 2025tie90.290.3tie
Questions people ask
Which is better, DeepSeek-V3.2 or Qwen3.5 122B A10B?
Qwen3.5 122B A10B wins three of the six areas where both have results: agents, long documents and following instructions. DeepSeek-V3.2 wins coding and facts, and costs 81% less. They are level on reasoning.
Which is better for coding?
DeepSeek-V3.2. It wins 4 of the 8 coding tests both models report; Qwen3.5 122B A10B wins 1, and 3 are ties.
Which is cheaper?
DeepSeek-V3.2 costs $0.28 per million input tokens and $0.42 per million output tokens; Qwen3.5 122B A10B costs $0.40 and $3.20. That makes DeepSeek-V3.2 about 81% cheaper for the same work.
How do you compare the two?
We use the 35 benchmark tests both models have published scores on. The verdict counts the 20 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 15 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.