Model comparison
DeepSeek V4 Pro vs QwQ-32B
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 39.8 on the Noometry Index.
Last verified . 23 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 23 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 7 categories and QwQ-32B in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 23.7.
- The biggest single-benchmark swing is Chess Puzzles: 47% for DeepSeek V4 Pro and 5% for QwQ-32B.
Side by side
| DeepSeek V4 Pro | QwQ-32B | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 54.3 | 39.8 |
| Released | 2026-04-24 | 2024-11-28 |
| Weights | Open | Open |
| Context window | 1M | — |
| Max output | 393K | — |
| Input $ / M tokens | $0.66 | — |
| Output $ / M tokens | $1.98 | — |
| Results tracked | 48 | 36 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), QwQ-32B: 35.4 (#226)
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| LMArena Coding | 1470 | 1333 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| Aider Polyglot | — | 20.9% |
| LMArena WebDev | 1582 | — |
| SciCode | 51% | — |
| WeirdML | 66.2% | — |
| BigCodeBench Instruct | — | 44.6% |
| LiveBench Coding | — | 72.2% |
| BigCodeBench Complete | — | 54.4% |
| ALE-Bench | 1,403 | — |
Agentic & Tool Use Not comparable
DeepSeek V4 Pro: 32.8 (#58), QwQ-32B: —
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| APEX-Agents | 47.3% | — |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), QwQ-32B: 23.7 (#174)
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| Chess Puzzles | 47% | 5% |
| LMArena Hard Prompts | 1461 | 1325 |
| Epoch Capabilities Index | 155.31 | 137.6 |
| ForecastBench | 56.1 | 58.3 |
| ARC-AGI-2 | 61.3% | — |
| Kagi LLM Benchmark | 53.5% | — |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 90.5% | — |
| CritPt | 18% | — |
| LiveBench Reasoning | — | 83.5% |
| Mystery Game Puzzles | 43% | — |
| DTBench | 93.9% | — |
| LiveBench Data Analysis | — | 65% |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| LiveBench | — | 72% |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), QwQ-32B: 38.0 (#143)
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 59.2% |
| LMArena Math | 1455 | 1359 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| ProofBench | 50% | — |
| LiveBench Math | — | 77.8% |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), QwQ-32B: 37.2 (#158)
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| GPQA Diamond | 91.7% | 65.3% |
| LMArena Expert | 1464 | 1324 |
| SimpleQA Verified | 52.9% | — |
| Confabulations | — | 15.6% |
| Vectara Hallucination Rate | 8.6% | — |
Multilingual DeepSeek V4 Pro leads
DeepSeek V4 Pro: 54.4 (#45), QwQ-32B: 44.8 (#176)
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| LMArena Non-English | 1439 | 1305 |
| LMArena Chinese | 1486 | 1378 |
| LMArena French | 1472 | 1336 |
| LMArena German | 1458 | 1313 |
| LMArena Japanese | 1445 | 1262 |
| LMArena Korean | 1447 | 1279 |
| LMArena Russian | 1453 | 1297 |
| LMArena Spanish | 1458 | 1354 |
Instruction Following DeepSeek V4 Pro leads
DeepSeek V4 Pro: 76.1 (#47), QwQ-32B: 72.6 (#137)
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| LMArena Instruction Following | 1448 | 1297 |
| LiveBench Instruction Following | — | 81.8% |
Long Context QwQ-32B leads
DeepSeek V4 Pro: 45.0 (#51), QwQ-32B: 49.0 (#11)
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| LMArena Longer Query | 1458 | 1308 |
| Fiction.LiveBench | — | 83.3% |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), QwQ-32B: 50.6 (#180)
| Benchmark | DeepSeek V4 Pro | QwQ-32B |
|---|---|---|
| LMArena Text | 1451 | 1329 |
| LMArena Creative Writing | 1446 | 1288 |
| EQ-Bench Creative Writing | 1553 | 1257 |
| LMArena Multi-Turn | 1467 | 1314 |
| Short-Story Creative Writing | — | 80.2% |
| EQ-Bench 4 | 1166 | — |
| LiveBench Language | — | 51.4% |
Frequently asked questions
Is DeepSeek V4 Pro better than QwQ-32B?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 39.8 on the Noometry Index.
Is DeepSeek V4 Pro or QwQ-32B better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 35.4 in the Noometry coding category.
How many benchmarks do DeepSeek V4 Pro and QwQ-32B share?
23 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and QwQ-32B has 36.