Model comparison
DeepSeek-V2.5 (Sep 2024) vs Qwen2.5 72B Instruct
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 31.9 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 7 categories and Qwen2.5 72B Instruct in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V2.5 (Sep 2024) leads 35.9 to 19.3.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 37.6 | 31.9 |
| Released | 2024-09-06 | 2024-09 |
| Weights | Open | Open |
| Context window | — | 131K |
| Max output | — | 8K |
| Input $ / M tokens | — | $1.40 |
| Output $ / M tokens | — | $5.60 |
| Results tracked | 22 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen2.5 72B Instruct leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Qwen2.5 72B Instruct: 33.2 (#260)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| BigCodeBench Instruct | 48.6% | 45.8% |
| LMArena Coding | 1309 | 1292 |
| BigCodeBench Complete | 53.2% | 55.9% |
| Aider Polyglot | 17.8% | — |
| WeirdML | — | 16% |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Qwen2.5 72B Instruct: 22.1 (#133)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| METR Time Horizons | — | 35.8% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Qwen2.5 72B Instruct: 22.3 (#199)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1271 |
| DTBench | — | 62.9% |
| LMCA | — | 13.4% |
| BIG-Bench Hard | — | 79.8% |
| Epoch Capabilities Index | — | 129 |
| ForecastBench | — | 57.5 |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Qwen2.5 72B Instruct: 19.3 (#287)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Math | 1288 | 1283 |
| OTIS Mock AIME 2024-2025 | — | 8.1% |
| Omni-MATH | — | 33% |
| MATH Level 5 | — | 63.2% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Qwen2.5 72B Instruct: 27.0 (#253)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Expert | 1266 | 1245 |
| GPQA Diamond | — | 49.1% |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| GPQA (HELM) | — | 42.6% |
| ARC (AI2) Challenge | — | 94.5% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Qwen2.5 72B Instruct: 41.0 (#213)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1273 | 1252 |
| LMArena Chinese | 1318 | 1272 |
| LMArena French | 1289 | 1280 |
| LMArena German | 1258 | 1234 |
| LMArena Japanese | 1228 | 1180 |
| LMArena Korean | 1209 | 1188 |
| LMArena Russian | 1289 | 1264 |
| LMArena Spanish | 1248 | 1256 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Qwen2.5 72B Instruct: 65.5 (#221)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1280 | 1254 |
| IFEval | — | 80.6% |
Long Context Too close to call
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Qwen2.5 72B Instruct: 38.9 (#188)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1301 | 1282 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Qwen2.5 72B Instruct: 46.7 (#215)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1294 | 1269 |
| LMArena Creative Writing | 1285 | 1221 |
| LMArena Multi-Turn | 1297 | 1272 |
| WildBench | — | 80.2% |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Qwen2.5 72B Instruct?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 31.9 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Qwen2.5 72B Instruct better for coding?
Qwen2.5 72B Instruct scores higher on coding benchmarks: 33.2 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Qwen2.5 72B Instruct share?
19 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Qwen2.5 72B Instruct has 43.