Model comparison
DeepSeek LLM 67B vs GPT-4
GPT-4 is the stronger model overall, scoring 29.1 to 24.9 on the Noometry Index.
Last verified . 15 shared benchmarks.
DeepSeekSummary
- They share 15 benchmarks with published results for both. DeepSeek LLM 67B scores higher in 1 category and GPT-4 in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where GPT-4 leads 18.4 to 7.0.
- The biggest single-benchmark swing is MATH Level 5: 6.4% for DeepSeek LLM 67B and 23% for GPT-4.
- DeepSeek LLM 67B has downloadable open weights; the other is API-only.
Side by side
| DeepSeek LLM 67B | GPT-4 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 24.9 | 29.1 |
| Released | 2023-11-29 | 2023-03-14 |
| Weights | Open | Proprietary |
| Context window | — | 8K |
| Max output | — | 8K |
| Input $ / M tokens | — | $30 |
| Output $ / M tokens | — | $60 |
| Results tracked | 15 | 38 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek LLM 67B: 31.9 (#278), GPT-4: 31.6 (#283)
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| LMArena Coding | 1096 | 1254 |
| WeirdML | — | 12.4% |
| BigCodeBench Instruct | — | 46% |
| BigCodeBench Complete | — | 57.2% |
| HumanEval+ | — | 79.3% |
Agentic & Tool Use Not comparable
DeepSeek LLM 67B: —, GPT-4: —
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| METR Time Horizons | — | 36.1% |
Reasoning GPT-4 leads
DeepSeek LLM 67B: 16.5 (#304), GPT-4: 17.8 (#289)
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| Chess Puzzles | 0% | 4% |
| LMArena Hard Prompts | 1070 | 1241 |
| Epoch Capabilities Index | 110.5 | 125.89 |
| Mystery Game Puzzles | — | 12% |
| DTBench | — | 62.7% |
| LMCA | — | 17.1% |
| BIG-Bench Hard | — | 75.1% |
| ForecastBench | — | 57.8 |
| HellaSwag | — | 95.3% |
| WinoGrande | — | 87.5% |
Math GPT-4 leads
DeepSeek LLM 67B: 8.7 (#324), GPT-4: 10.8 (#309)
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 0.8% | 1.1% |
| LMArena Math | 1108 | 1269 |
| MATH Level 5 | 6.4% | 23% |
| GSM8K | — | 92% |
Knowledge GPT-4 leads
DeepSeek LLM 67B: 7.0 (#313), GPT-4: 18.4 (#282)
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| GPQA Diamond | 24.6% | 35.7% |
| LMArena Expert | — | 1211 |
| MMLU | — | 86.4% |
| TriviaQA | — | 84.8% |
Multilingual GPT-4 leads
DeepSeek LLM 67B: 29.4 (#267), GPT-4: 40.6 (#215)
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| LMArena Non-English | 1073 | 1246 |
| LMArena Chinese | 1132 | 1242 |
| LMArena French | — | 1283 |
| LMArena German | — | 1251 |
| LMArena Japanese | — | 1209 |
| LMArena Korean | — | 1184 |
| LMArena Russian | — | 1251 |
| LMArena Spanish | — | 1261 |
Instruction Following GPT-4 leads
DeepSeek LLM 67B: 55.4 (#277), GPT-4: 65.3 (#222)
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| LMArena Instruction Following | 1079 | 1241 |
Long Context GPT-4 leads
DeepSeek LLM 67B: 33.1 (#265), GPT-4: 37.7 (#212)
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| LMArena Longer Query | 1092 | 1244 |
Writing & Preference GPT-4 leads
DeepSeek LLM 67B: 31.6 (#282), GPT-4: 34.9 (#268)
| Benchmark | DeepSeek LLM 67B | GPT-4 |
|---|---|---|
| LMArena Text | 1105 | 1263 |
| LMArena Creative Writing | 1067 | 1244 |
| LMArena Multi-Turn | 1082 | 1257 |
| EQ-Bench Creative Writing | — | 752 |
Frequently asked questions
Is DeepSeek LLM 67B better than GPT-4?
GPT-4 is the stronger model overall, scoring 29.1 to 24.9 on the Noometry Index.
Is DeepSeek LLM 67B or GPT-4 better for coding?
They score almost the same on coding (31.9 vs 31.6); test both on your own repository before choosing.
How many benchmarks do DeepSeek LLM 67B and GPT-4 share?
15 benchmarks have published results for both models. DeepSeek LLM 67B has 15 scored results on Noometry and GPT-4 has 38.