Model comparison
Deepseek Coder v2 vs DeepSeek-V2.5 (Sep 2024)
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 35.9 on the Noometry Index.
Last verified . 21 shared benchmarks.
DeepSeekSummary
- They share 21 benchmarks with published results for both. Deepseek Coder v2 scores higher in 1 category and DeepSeek-V2.5 (Sep 2024) in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-V2.5 (Sep 2024) leads 49.8 to 38.2.
- The biggest single-benchmark swing is BigCodeBench Complete: 59.7% for Deepseek Coder v2 and 53.2% for DeepSeek-V2.5 (Sep 2024).
Side by side
| Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 35.9 | 37.6 |
| Released | 2024-06-17 | 2024-09-06 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 24 | 22 |
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), DeepSeek-V2.5 (Sep 2024): 31.7 (#281)
| Benchmark | Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| BigCodeBench Instruct | 48.2% | 48.6% |
| LMArena Coding | 1251 | 1309 |
| BigCodeBench Complete | 59.7% | 53.2% |
| HumanEval+ | 82.3% | 83.5% |
| MBPP+ | 75.1% | 74.1% |
| Aider Polyglot | — | 17.8% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
Deepseek Coder v2: 23.6 (#176), DeepSeek-V2.5 (Sep 2024): 25.6 (#145)
| Benchmark | Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Hard Prompts | 1207 | 1289 |
| WinoGrande | 83.7% | — |
Math DeepSeek-V2.5 (Sep 2024) leads
Deepseek Coder v2: 34.9 (#190), DeepSeek-V2.5 (Sep 2024): 35.9 (#177)
| Benchmark | Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Math | 1241 | 1288 |
| GSM8K | 94.5% | — |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
Deepseek Coder v2: 32.3 (#212), DeepSeek-V2.5 (Sep 2024): 34.8 (#193)
| Benchmark | Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Expert | 1181 | 1266 |
| ARC (AI2) Challenge | 64.3% | — |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
Deepseek Coder v2: 36.3 (#240), DeepSeek-V2.5 (Sep 2024): 42.5 (#193)
| Benchmark | Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Non-English | 1182 | 1273 |
| LMArena Chinese | 1201 | 1318 |
| LMArena French | 1185 | 1289 |
| LMArena German | 1164 | 1258 |
| LMArena Japanese | 1126 | 1228 |
| LMArena Korean | 1104 | 1209 |
| LMArena Russian | 1188 | 1289 |
| LMArena Spanish | 1153 | 1248 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
Deepseek Coder v2: 61.7 (#242), DeepSeek-V2.5 (Sep 2024): 67.5 (#194)
| Benchmark | Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Instruction Following | 1180 | 1280 |
Long Context DeepSeek-V2.5 (Sep 2024) leads
Deepseek Coder v2: 37.0 (#224), DeepSeek-V2.5 (Sep 2024): 39.5 (#174)
| Benchmark | Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Longer Query | 1219 | 1301 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
Deepseek Coder v2: 38.2 (#253), DeepSeek-V2.5 (Sep 2024): 49.8 (#187)
| Benchmark | Deepseek Coder v2 | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Text | 1191 | 1294 |
| LMArena Creative Writing | 1120 | 1285 |
| LMArena Multi-Turn | 1177 | 1297 |
Frequently asked questions
Is Deepseek Coder v2 better than DeepSeek-V2.5 (Sep 2024)?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 35.9 on the Noometry Index.
Is Deepseek Coder v2 or DeepSeek-V2.5 (Sep 2024) better for coding?
Deepseek Coder v2 scores higher on coding benchmarks: 38.1 versus 31.7 in the Noometry coding category.
How many benchmarks do Deepseek Coder v2 and DeepSeek-V2.5 (Sep 2024) share?
21 benchmarks have published results for both models. Deepseek Coder v2 has 24 scored results on Noometry and DeepSeek-V2.5 (Sep 2024) has 22.