Model comparison
DeepSeek-V3.1 vs GPT-4.1
DeepSeek-V3.1 is the stronger model overall, scoring 42.8 to 35.9 on the Noometry Index.
Last verified . 27 shared benchmarks.
DeepSeekSummary
- They share 27 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 7 categories and GPT-4.1 in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3.1 leads 38.9 to 22.3.
- The biggest single-benchmark swing is DTBench: 82.7% for DeepSeek-V3.1 and 68.3% for GPT-4.1.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $2 / $8 for GPT-4.1.
- GPT-4.1 accepts more context: 1.05M tokens versus 164K.
- DeepSeek-V3.1 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.1 | GPT-4.1 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.8 | 35.9 |
| Released | 2025-08-21 | 2025-04-14 |
| Weights | Open | Proprietary |
| Context window | 164K | 1.05M |
| Max output | 8K | 33K |
| Input $ / M tokens | $0.25 | $2 |
| Output $ / M tokens | $0.95 | $8 |
| Results tracked | 27 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.1 leads
DeepSeek-V3.1: 40.3 (#144), GPT-4.1: 34.4 (#238)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| WeirdML | 38.4% | 39% |
| LMArena Coding | 1417 | 1391 |
| SWE-bench Verified | — | 48.5% |
| SWE-bench Verified (bash only) | — | 39.6% |
| Aider Polyglot | — | 52.4% |
| CadEval | — | 42% |
| ALE-Bench | — | 558.1 |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, GPT-4.1: 34.7 (#43)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 54% |
Reasoning DeepSeek-V3.1 leads
DeepSeek-V3.1: 27.9 (#110), GPT-4.1: 11.7 (#339)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| SimpleBench | 40% | 27% |
| Kagi LLM Benchmark | 53.2% | 52.3% |
| LMArena Hard Prompts | 1417 | 1384 |
| DTBench | 82.7% | 68.3% |
| LMCA | 24.3% | 25.6% |
| Epoch Capabilities Index | 139.92 | 136.78 |
| ForecastBench | 58 | 61.5 |
| ARC-AGI-2 | — | 0.4% |
| ARC-AGI-1 | — | 5.5% |
| Chess Puzzles | — | 6% |
| EnigmaEval | — | 2.2% |
Math DeepSeek-V3.1 leads
DeepSeek-V3.1: 38.9 (#122), GPT-4.1: 22.3 (#280)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| LMArena Math | 1420 | 1370 |
| FrontierMath (Tiers 1-3) | — | 6% |
| OTIS Mock AIME 2024-2025 | — | 38.3% |
| Omni-MATH | — | 47.1% |
| MATH Level 5 | — | 83% |
| FrontierMath (Feb 2025 set) | — | 5.5% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge DeepSeek-V3.1 leads
DeepSeek-V3.1: 43.7 (#90), GPT-4.1: 37.1 (#160)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| Vectara Hallucination Rate | 5.5% | 5.6% |
| LMArena Expert | 1405 | 1364 |
| GPQA Diamond | — | 66.9% |
| Humanity's Last Exam | — | 5.4% |
| SimpleQA Verified | — | 31.1% |
| MMLU-Pro | — | 81.1% |
| GPQA (HELM) | — | 65.9% |
Multimodal Not comparable
DeepSeek-V3.1: —, GPT-4.1: 38.2 (#67)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| LMArena Vision | — | 1211 |
| GeoBench | — | 72% |
Multilingual DeepSeek-V3.1 leads
DeepSeek-V3.1: 51.6 (#106), GPT-4.1: 49.4 (#133)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| LMArena Non-English | 1400 | 1370 |
| LMArena Chinese | 1469 | 1382 |
| LMArena French | 1447 | 1382 |
| LMArena German | 1411 | 1381 |
| LMArena Japanese | 1378 | 1319 |
| LMArena Korean | 1337 | 1339 |
| LMArena Russian | 1405 | 1377 |
| LMArena Spanish | 1431 | 1376 |
Instruction Following DeepSeek-V3.1 leads
DeepSeek-V3.1: 73.9 (#110), GPT-4.1: 71.3 (#153)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| LMArena Instruction Following | 1400 | 1367 |
| IFEval | — | 83.8% |
Long Context GPT-4.1 leads
DeepSeek-V3.1: 36.3 (#232), GPT-4.1: 40.0 (#163)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| Fiction.LiveBench | 52.8% | 63.9% |
| LMArena Longer Query | 1422 | 1385 |
Writing & Preference DeepSeek-V3.1 leads
DeepSeek-V3.1: 60.3 (#98), GPT-4.1: 57.6 (#125)
| Benchmark | DeepSeek-V3.1 | GPT-4.1 |
|---|---|---|
| LMArena Text | 1420 | 1383 |
| LMArena Creative Writing | 1401 | 1363 |
| EQ-Bench Creative Writing | 1436 | 1420 |
| LMArena Multi-Turn | 1408 | 1398 |
| WildBench | — | 85.4% |
Frequently asked questions
Is DeepSeek-V3.1 better than GPT-4.1?
DeepSeek-V3.1 is the stronger model overall, scoring 42.8 to 35.9 on the Noometry Index.
Which is cheaper, DeepSeek-V3.1 or GPT-4.1?
DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; GPT-4.1 lists at $2 and $8.
Is DeepSeek-V3.1 or GPT-4.1 better for coding?
DeepSeek-V3.1 scores higher on coding benchmarks: 40.3 versus 34.4 in the Noometry coding category.
Which has the bigger context window?
GPT-4.1 does, with 1.05M tokens against 164K.
How many benchmarks do DeepSeek-V3.1 and GPT-4.1 share?
27 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and GPT-4.1 has 52.