Model comparison
DeepSeek-V2.5 (Sep 2024) vs GPT-4 Turbo
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 30.5 on the Noometry Index.
Last verified . 21 shared benchmarks.
DeepSeekSummary
- They share 21 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 7 categories and GPT-4 Turbo in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V2.5 (Sep 2024) leads 35.9 to 9.0.
- The biggest single-benchmark swing is BigCodeBench Complete: 53.2% for DeepSeek-V2.5 (Sep 2024) and 58.2% for GPT-4 Turbo.
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 37.6 | 30.5 |
| Released | 2024-09-06 | 2023-11-06 |
| Weights | Open | Proprietary |
| Context window | — | 128K |
| Max output | — | 4K |
| Input $ / M tokens | — | $10 |
| Output $ / M tokens | — | $30 |
| Results tracked | 22 | 36 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-4 Turbo leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), GPT-4 Turbo: 33.8 (#249)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| BigCodeBench Instruct | 48.6% | 48.2% |
| LMArena Coding | 1309 | 1268 |
| BigCodeBench Complete | 53.2% | 58.2% |
| HumanEval+ | 83.5% | 86.6% |
| MBPP+ | 74.1% | 73.3% |
| Aider Polyglot | 17.8% | — |
| WeirdML | — | 18% |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, GPT-4 Turbo: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| METR Time Horizons | — | 36.7% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), GPT-4 Turbo: 15.3 (#317)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1251 |
| SimpleBench | — | 25.1% |
| Chess Puzzles | — | 6% |
| DTBench | — | 61.6% |
| LMCA | — | 9.8% |
| Epoch Capabilities Index | — | 127.25 |
| ForecastBench | — | 59.4 |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), GPT-4 Turbo: 9.0 (#322)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| LMArena Math | 1288 | 1272 |
| FrontierMath (Tiers 1-3) | — | 0.7% |
| OTIS Mock AIME 2024-2025 | — | 6.7% |
| MATH Level 5 | — | 46.7% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), GPT-4 Turbo: 24.3 (#268)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| LMArena Expert | 1266 | 1223 |
| GPQA Diamond | — | 46.6% |
| Confabulations | — | 28.4% |
| MMLU | — | 81.3% |
Multimodal Not comparable
DeepSeek-V2.5 (Sep 2024): —, GPT-4 Turbo: 30.6 (#110)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| LMArena Vision | — | 1090 |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), GPT-4 Turbo: 40.5 (#216)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| LMArena Non-English | 1273 | 1245 |
| LMArena Chinese | 1318 | 1242 |
| LMArena French | 1289 | 1276 |
| LMArena German | 1258 | 1259 |
| LMArena Japanese | 1228 | 1194 |
| LMArena Korean | 1209 | 1187 |
| LMArena Russian | 1289 | 1259 |
| LMArena Spanish | 1248 | 1260 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), GPT-4 Turbo: 65.8 (#216)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| LMArena Instruction Following | 1280 | 1249 |
Long Context DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), GPT-4 Turbo: 38.0 (#206)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| LMArena Longer Query | 1301 | 1254 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), GPT-4 Turbo: 47.7 (#206)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | GPT-4 Turbo |
|---|---|---|
| LMArena Text | 1294 | 1272 |
| LMArena Creative Writing | 1285 | 1269 |
| LMArena Multi-Turn | 1297 | 1267 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than GPT-4 Turbo?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 30.5 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or GPT-4 Turbo better for coding?
GPT-4 Turbo scores higher on coding benchmarks: 33.8 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and GPT-4 Turbo share?
21 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and GPT-4 Turbo has 36.