Model comparison
DeepSeek-V3.2-Speciale vs Qwen3 235B-A22B
Qwen3 235B-A22B is the stronger model overall, scoring 43.5 to 39.7 on the Noometry Index.
Last verified . 3 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 3 benchmarks with published results for both. DeepSeek-V3.2-Speciale scores higher in 1 category and Qwen3 235B-A22B in 2 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek-V3.2-Speciale leads 32.9 to 15.7.
- The biggest single-benchmark swing is SimpleBench: 52.6% for DeepSeek-V3.2-Speciale and 31% for Qwen3 235B-A22B.
- DeepSeek-V3.2-Speciale is cheaper at $0.58 / $1.68 per million input/output tokens, against $0.70 / $2.80 for Qwen3 235B-A22B.
- Qwen3 235B-A22B accepts more context: 131K tokens versus 128K.
Side by side
| DeepSeek-V3.2-Speciale | Qwen3 235B-A22B | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 39.7 | 43.5 |
| Released | 2025-12-01 | 2025-04 |
| Weights | Open | Open |
| Context window | 128K | 131K |
| Max output | 128K | 16K |
| Input $ / M tokens | $0.58 | $0.70 |
| Output $ / M tokens | $1.68 | $2.80 |
| Results tracked | 3 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3 235B-A22B leads
DeepSeek-V3.2-Speciale: 40.4 (#140), Qwen3 235B-A22B: 44.3 (#75)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| WeirdML | 46.7% | 41% |
| Aider Polyglot | — | 59.6% |
| SciCode | — | 42.4% |
| LMArena Coding | — | 1445 |
Agentic & Tool Use Not comparable
DeepSeek-V3.2-Speciale: —, Qwen3 235B-A22B: 33.9 (#51)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 52.1% |
| Vending-Bench 2 | — | -11.34 |
Reasoning DeepSeek-V3.2-Speciale leads
DeepSeek-V3.2-Speciale: 32.9 (#73), Qwen3 235B-A22B: 15.7 (#311)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| SimpleBench | 52.6% | 31% |
| ARC-AGI-2 | — | 1.3% |
| Kagi LLM Benchmark | — | 69.4% |
| ARC-AGI-1 | — | 11% |
| CritPt | — | 0% |
| Chess Puzzles | — | 12% |
| LMArena Hard Prompts | — | 1433 |
| Mystery Game Puzzles | — | 9% |
| DTBench | — | 80.3% |
| LMCA | — | 29.3% |
| Epoch Capabilities Index | — | 143.85 |
| ForecastBench | — | 59.7 |
Math Not comparable
DeepSeek-V3.2-Speciale: —, Qwen3 235B-A22B: 50.4 (#57)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | — | 86.7% |
| Omni-MATH | — | 71.8% |
| LMArena Math | — | 1432 |
| MATH Level 5 | — | 68.9% |
| FrontierMath (Feb 2025 set) | — | 8.5% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge Not comparable
DeepSeek-V3.2-Speciale: —, Qwen3 235B-A22B: 49.6 (#73)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| GPQA Diamond | — | 80.1% |
| SimpleQA Verified | — | 40.4% |
| MMLU-Pro | — | 84.4% |
| Confabulations | — | 15.6% |
| Vectara Hallucination Rate | — | 9.3% |
| GPQA (HELM) | — | 72.7% |
| LMArena Expert | — | 1463 |
Multilingual Not comparable
DeepSeek-V3.2-Speciale: —, Qwen3 235B-A22B: 52.3 (#89)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| LMArena Non-English | — | 1409 |
| LMArena Chinese | — | 1481 |
| LMArena French | — | 1445 |
| LMArena German | — | 1433 |
| LMArena Japanese | — | 1399 |
| LMArena Korean | — | 1391 |
| LMArena Russian | — | 1411 |
| LMArena Spanish | — | 1430 |
Instruction Following Not comparable
DeepSeek-V3.2-Speciale: —, Qwen3 235B-A22B: 72.6 (#136)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| IFEval | — | 83.5% |
| LMArena Instruction Following | — | 1408 |
Long Context Not comparable
DeepSeek-V3.2-Speciale: —, Qwen3 235B-A22B: 46.1 (#26)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| Fiction.LiveBench | — | 75% |
| LMArena Longer Query | — | 1426 |
Writing & Preference Qwen3 235B-A22B leads
DeepSeek-V3.2-Speciale: 46.0 (#222), Qwen3 235B-A22B: 59.6 (#108)
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3 235B-A22B |
|---|---|---|
| EQ-Bench Creative Writing | 1276 | 1366 |
| LMArena Text | — | 1419 |
| LMArena Creative Writing | — | 1384 |
| Short-Story Creative Writing | — | 83% |
| WildBench | — | 86.6% |
| LMArena Multi-Turn | — | 1432 |
Frequently asked questions
Is DeepSeek-V3.2-Speciale better than Qwen3 235B-A22B?
Qwen3 235B-A22B is the stronger model overall, scoring 43.5 to 39.7 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Speciale or Qwen3 235B-A22B?
DeepSeek-V3.2-Speciale is cheaper. It lists at $0.58 per million input tokens and $1.68 per million output tokens; Qwen3 235B-A22B lists at $0.70 and $2.80.
Is DeepSeek-V3.2-Speciale or Qwen3 235B-A22B better for coding?
Qwen3 235B-A22B scores higher on coding benchmarks: 44.3 versus 40.4 in the Noometry coding category.
Which has the bigger context window?
Qwen3 235B-A22B does, with 131K tokens against 128K.
How many benchmarks do DeepSeek-V3.2-Speciale and Qwen3 235B-A22B share?
3 benchmarks have published results for both models. DeepSeek-V3.2-Speciale has 3 scored results on Noometry and Qwen3 235B-A22B has 49.