Model comparison
DeepSeek-V3.1 vs GPT-4 Turbo
DeepSeek-V3.1 is the stronger model overall, scoring 42.2 to 30.3 on the Noometry Index.
Last verified . 23 shared benchmarks.
DeepSeekSummary
- They share 23 benchmarks with published results for both. DeepSeek-V3.1 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-V3.1 leads 36.1 to 8.6.
- The biggest single-benchmark swing is DTBench: 82.7% for DeepSeek-V3.1 and 61.6% for GPT-4 Turbo.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $10 / $30 for GPT-4 Turbo.
- DeepSeek-V3.1 accepts more context: 164K tokens versus 128K.
- DeepSeek-V3.1 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.1 | GPT-4 Turbo | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.2 | 30.3 |
| Released | 2025-08-21 | 2023-11-06 |
| Weights | Open | Proprietary |
| Context window | 164K | 128K |
| Max output | 8K | 4K |
| Input $ / M tokens | $0.25 | $10 |
| Output $ / M tokens | $0.95 | $30 |
| Results tracked | 27 | 36 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.1 leads
DeepSeek-V3.1: 39.8 (#140), GPT-4 Turbo: 33.4 (#247)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| WeirdML | 38.4% | 18% |
| LMArena Coding | 1417 | 1268 |
| BigCodeBench Instruct | — | 48.2% |
| BigCodeBench Complete | — | 58.2% |
| HumanEval+ | — | 86.6% |
| MBPP+ | — | 73.3% |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, GPT-4 Turbo: —
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| METR Time Horizons | — | 36.7% |
Reasoning DeepSeek-V3.1 leads
DeepSeek-V3.1: 27.3 (#109), GPT-4 Turbo: 14.9 (#311)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| SimpleBench | 40% | 25.1% |
| LMArena Hard Prompts | 1417 | 1251 |
| DTBench | 82.7% | 61.6% |
| LMCA | 24.3% | 9.8% |
| Epoch Capabilities Index | 139.92 | 127.25 |
| ForecastBench | 58 | 59.4 |
| Kagi LLM Benchmark | 53.2% | — |
| Chess Puzzles | — | 6% |
Math DeepSeek-V3.1 leads
DeepSeek-V3.1: 36.1 (#125), GPT-4 Turbo: 8.6 (#316)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| LMArena Math | 1420 | 1272 |
| FrontierMath (Tiers 1-3) | — | 0.7% |
| OTIS Mock AIME 2024-2025 | — | 6.7% |
| MATH Level 5 | — | 46.7% |
Knowledge DeepSeek-V3.1 leads
DeepSeek-V3.1: 43.0 (#86), GPT-4 Turbo: 24.3 (#263)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| LMArena Expert | 1403 | 1223 |
| GPQA Diamond | — | 46.6% |
| Confabulations | — | 28.4% |
| Vectara Hallucination Rate | 5.5% | — |
| MMLU | — | 81.3% |
Multimodal Not comparable
DeepSeek-V3.1: —, GPT-4 Turbo: 30.0 (#104)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| LMArena Vision | — | 1090 |
Multilingual DeepSeek-V3.1 leads
DeepSeek-V3.1: 51.8 (#100), GPT-4 Turbo: 40.7 (#211)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| LMArena Non-English | 1400 | 1245 |
| LMArena Chinese | 1470 | 1242 |
| LMArena French | 1447 | 1275 |
| LMArena German | 1412 | 1259 |
| LMArena Japanese | 1374 | 1194 |
| LMArena Korean | 1337 | 1187 |
| LMArena Russian | 1405 | 1259 |
| LMArena Spanish | 1429 | 1260 |
Instruction Following DeepSeek-V3.1 leads
DeepSeek-V3.1: 74.1 (#105), GPT-4 Turbo: 66.0 (#211)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| LMArena Instruction Following | 1400 | 1250 |
Long Context GPT-4 Turbo leads
DeepSeek-V3.1: 36.3 (#227), GPT-4 Turbo: 38.0 (#200)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| LMArena Longer Query | 1422 | 1254 |
| Fiction.LiveBench | 52.8% | — |
Writing & Preference DeepSeek-V3.1 leads
DeepSeek-V3.1: 60.5 (#93), GPT-4 Turbo: 47.9 (#201)
| Benchmark | DeepSeek-V3.1 | GPT-4 Turbo |
|---|---|---|
| LMArena Text | 1420 | 1272 |
| LMArena Creative Writing | 1401 | 1269 |
| LMArena Multi-Turn | 1408 | 1267 |
| EQ-Bench Creative Writing | 1436 | — |
Frequently asked questions
Is DeepSeek-V3.1 better than GPT-4 Turbo?
DeepSeek-V3.1 is the stronger model overall, scoring 42.2 to 30.3 on the Noometry Index.
Which is cheaper, DeepSeek-V3.1 or GPT-4 Turbo?
DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; GPT-4 Turbo lists at $10 and $30.
Is DeepSeek-V3.1 or GPT-4 Turbo better for coding?
DeepSeek-V3.1 scores higher on coding benchmarks: 39.8 versus 33.4 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3.1 does, with 164K tokens against 128K.
How many benchmarks do DeepSeek-V3.1 and GPT-4 Turbo share?
23 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and GPT-4 Turbo has 36.