Model comparison
DeepSeek-V3.1 vs Qwen2.5 72B Instruct
DeepSeek-V3.1 is the stronger model overall, scoring 42.8 to 31.9 on the Noometry Index.
Last verified . 22 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 22 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 7 categories and Qwen2.5 72B Instruct in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3.1 leads 38.9 to 19.3.
- The biggest single-benchmark swing is WeirdML: 38.4% for DeepSeek-V3.1 and 16% for Qwen2.5 72B Instruct.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $1.40 / $5.60 for Qwen2.5 72B Instruct.
- DeepSeek-V3.1 accepts more context: 164K tokens versus 131K.
Side by side
| DeepSeek-V3.1 | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 42.8 | 31.9 |
| Released | 2025-08-21 | 2024-09 |
| Weights | Open | Open |
| Context window | 164K | 131K |
| Max output | 8K | 8K |
| Input $ / M tokens | $0.25 | $1.40 |
| Output $ / M tokens | $0.95 | $5.60 |
| Results tracked | 27 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.1 leads
DeepSeek-V3.1: 40.3 (#144), Qwen2.5 72B Instruct: 33.2 (#260)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| WeirdML | 38.4% | 16% |
| LMArena Coding | 1417 | 1292 |
| BigCodeBench Instruct | — | 45.8% |
| BigCodeBench Complete | — | 55.9% |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, Qwen2.5 72B Instruct: 22.1 (#133)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| METR Time Horizons | — | 35.8% |
Reasoning DeepSeek-V3.1 leads
DeepSeek-V3.1: 27.9 (#110), Qwen2.5 72B Instruct: 22.3 (#199)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1417 | 1271 |
| DTBench | 82.7% | 62.9% |
| LMCA | 24.3% | 13.4% |
| Epoch Capabilities Index | 139.92 | 129 |
| ForecastBench | 58 | 57.5 |
| SimpleBench | 40% | — |
| Kagi LLM Benchmark | 53.2% | — |
| BIG-Bench Hard | — | 79.8% |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math DeepSeek-V3.1 leads
DeepSeek-V3.1: 38.9 (#122), Qwen2.5 72B Instruct: 19.3 (#287)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Math | 1420 | 1283 |
| OTIS Mock AIME 2024-2025 | — | 8.1% |
| Omni-MATH | — | 33% |
| MATH Level 5 | — | 63.2% |
Knowledge DeepSeek-V3.1 leads
DeepSeek-V3.1: 43.7 (#90), Qwen2.5 72B Instruct: 27.0 (#253)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Expert | 1405 | 1245 |
| GPQA Diamond | — | 49.1% |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| Vectara Hallucination Rate | 5.5% | — |
| GPQA (HELM) | — | 42.6% |
| ARC (AI2) Challenge | — | 94.5% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |
Multilingual DeepSeek-V3.1 leads
DeepSeek-V3.1: 51.6 (#106), Qwen2.5 72B Instruct: 41.0 (#213)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1400 | 1252 |
| LMArena Chinese | 1469 | 1272 |
| LMArena French | 1447 | 1280 |
| LMArena German | 1411 | 1234 |
| LMArena Japanese | 1378 | 1180 |
| LMArena Korean | 1337 | 1188 |
| LMArena Russian | 1405 | 1264 |
| LMArena Spanish | 1431 | 1256 |
Instruction Following DeepSeek-V3.1 leads
DeepSeek-V3.1: 73.9 (#110), Qwen2.5 72B Instruct: 65.5 (#221)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1400 | 1254 |
| IFEval | — | 80.6% |
Long Context Qwen2.5 72B Instruct leads
DeepSeek-V3.1: 36.3 (#232), Qwen2.5 72B Instruct: 38.9 (#188)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1422 | 1282 |
| Fiction.LiveBench | 52.8% | — |
Writing & Preference DeepSeek-V3.1 leads
DeepSeek-V3.1: 60.3 (#98), Qwen2.5 72B Instruct: 46.7 (#215)
| Benchmark | DeepSeek-V3.1 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1420 | 1269 |
| LMArena Creative Writing | 1401 | 1221 |
| LMArena Multi-Turn | 1408 | 1272 |
| EQ-Bench Creative Writing | 1436 | — |
| WildBench | — | 80.2% |
Frequently asked questions
Is DeepSeek-V3.1 better than Qwen2.5 72B Instruct?
DeepSeek-V3.1 is the stronger model overall, scoring 42.8 to 31.9 on the Noometry Index.
Which is cheaper, DeepSeek-V3.1 or Qwen2.5 72B Instruct?
DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.
Is DeepSeek-V3.1 or Qwen2.5 72B Instruct better for coding?
DeepSeek-V3.1 scores higher on coding benchmarks: 40.3 versus 33.2 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3.1 does, with 164K tokens against 131K.
How many benchmarks do DeepSeek-V3.1 and Qwen2.5 72B Instruct share?
22 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and Qwen2.5 72B Instruct has 43.