Model comparison
DeepSeek-V3 vs Qwen2.5 32B Instruct
DeepSeek-V3 is the stronger model overall, scoring 39.2 to 29.7 on the Noometry Index.
Last verified . 6 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 6 benchmarks with published results for both. DeepSeek-V3 scores higher in 4 categories and Qwen2.5 32B Instruct in 0 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3 leads 30.4 to 15.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 7.4% for Qwen2.5 32B Instruct.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $0.70 / $2.80 for Qwen2.5 32B Instruct.
- DeepSeek-V3 accepts more context: 164K tokens versus 131K.
Side by side
| DeepSeek-V3 | Qwen2.5 32B Instruct | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 39.2 | 29.7 |
| Released | 2024-12-26 | 2024-09 |
| Weights | Open | Open |
| Context window | 164K | 131K |
| Max output | 164K | 8K |
| Input $ / M tokens | $0.24 | $0.70 |
| Output $ / M tokens | $0.90 | $2.80 |
| Results tracked | 60 | 7 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
DeepSeek-V3: 41.9 (#99), Qwen2.5 32B Instruct: 38.1 (#165)
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| BigCodeBench Instruct | 50% | 45% |
| BigCodeBench Complete | 62.2% | 52.3% |
| Aider Polyglot | 55.1% | — |
| SciCode | 35.8% | — |
| WeirdML | 36.1% | — |
| LiveBench Coding | 70.9% | — |
| LMArena Coding | 1368 | — |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| METR Time Horizons | 49.6% | — |
Reasoning DeepSeek-V3 leads
DeepSeek-V3: 20.0 (#226), Qwen2.5 32B Instruct: 18.5 (#262)
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| Epoch Capabilities Index | 135.95 | 128.52 |
| SimpleBench | 27.2% | — |
| Kagi LLM Benchmark | 52.3% | — |
| CritPt | 0% | — |
| Chess Puzzles | — | 0% |
| LiveBench Reasoning | 65.8% | — |
| LMArena Hard Prompts | 1365 | — |
| DTBench | 64.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| LMCA | 15.5% | — |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math DeepSeek-V3 leads
DeepSeek-V3: 30.4 (#208), Qwen2.5 32B Instruct: 15.3 (#289)
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 7.4% |
| MATH Level 5 | 75.5% | 56.1% |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| LMArena Math | 1373 | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
Knowledge DeepSeek-V3 leads
DeepSeek-V3: 37.5 (#147), Qwen2.5 32B Instruct: 24.9 (#259)
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| GPQA Diamond | 67.6% | 46.1% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| LMArena Expert | 1352 | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multilingual Not comparable
DeepSeek-V3: 48.8 (#138), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| LMArena Non-English | 1358 | — |
| LMArena Chinese | 1390 | — |
| LMArena French | 1386 | — |
| LMArena German | 1374 | — |
| LMArena Japanese | 1333 | — |
| LMArena Korean | 1321 | — |
| LMArena Russian | 1372 | — |
| LMArena Spanish | 1358 | — |
Instruction Following Not comparable
DeepSeek-V3: 73.0 (#125), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |
| LMArena Instruction Following | 1345 | — |
Long Context Not comparable
DeepSeek-V3: 34.0 (#248), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| Fiction.LiveBench | 50% | — |
| LMArena Longer Query | 1353 | — |
Writing & Preference Not comparable
DeepSeek-V3: 57.6 (#124), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek-V3 | Qwen2.5 32B Instruct |
|---|---|---|
| LMArena Text | 1375 | — |
| LMArena Creative Writing | 1364 | — |
| Short-Story Creative Writing | 77% | — |
| EQ-Bench Creative Writing | 1472 | — |
| WildBench | 83% | — |
| LMArena Multi-Turn | 1389 | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than Qwen2.5 32B Instruct?
DeepSeek-V3 is the stronger model overall, scoring 39.2 to 29.7 on the Noometry Index.
Which is cheaper, DeepSeek-V3 or Qwen2.5 32B Instruct?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Qwen2.5 32B Instruct lists at $0.70 and $2.80.
Is DeepSeek-V3 or Qwen2.5 32B Instruct better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 41.9 versus 38.1 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3 does, with 164K tokens against 131K.
How many benchmarks do DeepSeek-V3 and Qwen2.5 32B Instruct share?
6 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Qwen2.5 32B Instruct has 7.