Model comparison
Deepseek Coder v2 vs Qwen2-72B
Deepseek Coder v2 is the stronger model overall, scoring 35.9 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 Coder v2 scores higher in 6 categories and Qwen2-72B in 2 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Deepseek Coder v2 leads 32.3 to 21.2.
- The biggest single-benchmark swing is BigCodeBench Instruct: 48.2% for Deepseek Coder v2 and 38.5% for Qwen2-72B.
Side by side
| Deepseek Coder v2 | Qwen2-72B | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 35.9 | 30.0 |
| Released | 2024-06-17 | 2024-06-07 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 24 | 26 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Deepseek Coder v2 leads
Deepseek Coder v2: 38.1 (#183), Qwen2-72B: 29.1 (#310)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| BigCodeBench Instruct | 48.2% | 38.5% |
| LMArena Coding | 1251 | 1196 |
| BigCodeBench Complete | 59.7% | 54% |
| WeirdML | — | 11.3% |
| HumanEval+ | 82.3% | — |
| MBPP+ | 75.1% | — |
Agentic & Tool Use Not comparable
Deepseek Coder v2: —, Qwen2-72B: 17.0 (#146)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| TheAgentCompany | — | 1.1% |
| METR Time Horizons | — | 29.9% |
Reasoning Too close to call
Deepseek Coder v2: 23.6 (#176), Qwen2-72B: 23.2 (#181)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| LMArena Hard Prompts | 1207 | 1191 |
| Epoch Capabilities Index | — | 125.28 |
| WinoGrande | 83.7% | — |
Math Deepseek Coder v2 leads
Deepseek Coder v2: 34.9 (#190), Qwen2-72B: 30.2 (#236)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| LMArena Math | 1241 | 1235 |
| MATH Level 5 | — | 39.1% |
| GSM8K | 94.5% | — |
Knowledge Deepseek Coder v2 leads
Deepseek Coder v2: 32.3 (#212), Qwen2-72B: 21.2 (#275)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| LMArena Expert | 1181 | 1171 |
| GPQA Diamond | — | 40.8% |
| ARC (AI2) Challenge | 64.3% | — |
| MMLU | — | 82.4% |
Multilingual Too close to call
Deepseek Coder v2: 36.3 (#240), Qwen2-72B: 35.9 (#244)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| LMArena Non-English | 1182 | 1176 |
| LMArena Chinese | 1201 | 1240 |
| LMArena French | 1185 | 1170 |
| LMArena German | 1164 | 1151 |
| LMArena Japanese | 1126 | 1111 |
| LMArena Korean | 1104 | 1083 |
| LMArena Russian | 1188 | 1169 |
| LMArena Spanish | 1153 | 1169 |
Instruction Following Too close to call
Deepseek Coder v2: 61.7 (#242), Qwen2-72B: 61.7 (#241)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| LMArena Instruction Following | 1180 | 1181 |
Long Context Too close to call
Deepseek Coder v2: 37.0 (#224), Qwen2-72B: 36.1 (#235)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| LMArena Longer Query | 1219 | 1192 |
Writing & Preference Qwen2-72B leads
Deepseek Coder v2: 38.2 (#253), Qwen2-72B: 40.8 (#241)
| Benchmark | Deepseek Coder v2 | Qwen2-72B |
|---|---|---|
| LMArena Text | 1191 | 1203 |
| LMArena Creative Writing | 1120 | 1181 |
| LMArena Multi-Turn | 1177 | 1196 |
Frequently asked questions
Is Deepseek Coder v2 better than Qwen2-72B?
Deepseek Coder v2 is the stronger model overall, scoring 35.9 to 30.0 on the Noometry Index.
Is Deepseek Coder v2 or Qwen2-72B better for coding?
Deepseek Coder v2 scores higher on coding benchmarks: 38.1 versus 29.1 in the Noometry coding category.
How many benchmarks do Deepseek Coder v2 and Qwen2-72B share?
19 benchmarks have published results for both models. Deepseek Coder v2 has 24 scored results on Noometry and Qwen2-72B has 26.