Model comparison
DeepSeek-V2.5 (Sep 2024) vs Qwen2-72B
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 30.0 on the Noometry Index.
Last verified . 19 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 19 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 8 categories and Qwen2-72B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V2.5 (Sep 2024) leads 34.8 to 21.2.
- The biggest single-benchmark swing is BigCodeBench Instruct: 48.6% for DeepSeek-V2.5 (Sep 2024) and 38.5% for Qwen2-72B.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Qwen2-72B | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 37.6 | 30.0 |
| Released | 2024-09-06 | 2024-06-07 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 22 | 26 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Qwen2-72B: 29.1 (#310)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| BigCodeBench Instruct | 48.6% | 38.5% |
| LMArena Coding | 1309 | 1196 |
| BigCodeBench Complete | 53.2% | 54% |
| Aider Polyglot | 17.8% | — |
| WeirdML | — | 11.3% |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Qwen2-72B: 17.0 (#146)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| TheAgentCompany | — | 1.1% |
| METR Time Horizons | — | 29.9% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Qwen2-72B: 23.2 (#181)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1191 |
| Epoch Capabilities Index | — | 125.28 |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Qwen2-72B: 30.2 (#236)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| LMArena Math | 1288 | 1235 |
| MATH Level 5 | — | 39.1% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Qwen2-72B: 21.2 (#275)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| LMArena Expert | 1266 | 1171 |
| GPQA Diamond | — | 40.8% |
| MMLU | — | 82.4% |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Qwen2-72B: 35.9 (#244)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| LMArena Non-English | 1273 | 1176 |
| LMArena Chinese | 1318 | 1240 |
| LMArena French | 1289 | 1170 |
| LMArena German | 1258 | 1151 |
| LMArena Japanese | 1228 | 1111 |
| LMArena Korean | 1209 | 1083 |
| LMArena Russian | 1289 | 1169 |
| LMArena Spanish | 1248 | 1169 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Qwen2-72B: 61.7 (#241)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| LMArena Instruction Following | 1280 | 1181 |
Long Context DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Qwen2-72B: 36.1 (#235)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| LMArena Longer Query | 1301 | 1192 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Qwen2-72B: 40.8 (#241)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2-72B |
|---|---|---|
| LMArena Text | 1294 | 1203 |
| LMArena Creative Writing | 1285 | 1181 |
| LMArena Multi-Turn | 1297 | 1196 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Qwen2-72B?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 30.0 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Qwen2-72B better for coding?
DeepSeek-V2.5 (Sep 2024) scores higher on coding benchmarks: 31.7 versus 29.1 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Qwen2-72B share?
19 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Qwen2-72B has 26.