Model comparison
DeepSeek-V2.5 (Sep 2024) vs GPT-4
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 29.1 on the Noometry Index.
Last verified . 20 shared benchmarks.
DeepSeekSummary
- They share 20 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 8 categories and GPT-4 in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V2.5 (Sep 2024) leads 35.9 to 10.8.
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V2.5 (Sep 2024) | GPT-4 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 37.6 | 29.1 |
| Released | 2024-09-06 | 2023-03-14 |
| Weights | Open | Proprietary |
| Context window | — | 8K |
| Max output | — | 8K |
| Input $ / M tokens | — | $30 |
| Output $ / M tokens | — | $60 |
| Results tracked | 22 | 38 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), GPT-4: 31.6 (#283)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| BigCodeBench Instruct | 48.6% | 46% |
| LMArena Coding | 1309 | 1254 |
| BigCodeBench Complete | 53.2% | 57.2% |
| HumanEval+ | 83.5% | 79.3% |
| Aider Polyglot | 17.8% | — |
| WeirdML | — | 12.4% |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, GPT-4: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| METR Time Horizons | — | 36.1% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), GPT-4: 17.8 (#289)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1241 |
| Chess Puzzles | — | 4% |
| Mystery Game Puzzles | — | 12% |
| DTBench | — | 62.7% |
| LMCA | — | 17.1% |
| BIG-Bench Hard | — | 75.1% |
| Epoch Capabilities Index | — | 125.89 |
| ForecastBench | — | 57.8 |
| HellaSwag | — | 95.3% |
| WinoGrande | — | 87.5% |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), GPT-4: 10.8 (#309)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| LMArena Math | 1288 | 1269 |
| OTIS Mock AIME 2024-2025 | — | 1.1% |
| MATH Level 5 | — | 23% |
| GSM8K | — | 92% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), GPT-4: 18.4 (#282)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| LMArena Expert | 1266 | 1211 |
| GPQA Diamond | — | 35.7% |
| MMLU | — | 86.4% |
| TriviaQA | — | 84.8% |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), GPT-4: 40.6 (#215)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| LMArena Non-English | 1273 | 1246 |
| LMArena Chinese | 1318 | 1242 |
| LMArena French | 1289 | 1283 |
| LMArena German | 1258 | 1251 |
| LMArena Japanese | 1228 | 1209 |
| LMArena Korean | 1209 | 1184 |
| LMArena Russian | 1289 | 1251 |
| LMArena Spanish | 1248 | 1261 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), GPT-4: 65.3 (#222)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| LMArena Instruction Following | 1280 | 1241 |
Long Context DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), GPT-4: 37.7 (#212)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| LMArena Longer Query | 1301 | 1244 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), GPT-4: 34.9 (#268)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 |
|---|---|---|
| LMArena Text | 1294 | 1263 |
| LMArena Creative Writing | 1285 | 1244 |
| LMArena Multi-Turn | 1297 | 1257 |
| EQ-Bench Creative Writing | — | 752 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than GPT-4?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 29.1 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or GPT-4 better for coding?
They score almost the same on coding (31.7 vs 31.6); test both on your own repository before choosing.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and GPT-4 share?
20 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and GPT-4 has 38.