Model comparison
DeepSeek-V3 vs GPT-4.1
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 35.9 on the Noometry Index.
Last verified . 37 shared benchmarks.
DeepSeekSummary
- They share 37 benchmarks with published results for both. DeepSeek-V3 scores higher in 5 categories and GPT-4.1 in 3 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3 leads 32.1 to 22.3.
- The biggest single-benchmark swing is Fiction.LiveBench: 50% for DeepSeek-V3 and 63.9% for GPT-4.1.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 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 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3 | GPT-4.1 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 39.5 | 35.9 |
| Released | 2024-12-26 | 2025-04-14 |
| Weights | Open | Proprietary |
| Context window | 164K | 1.05M |
| Max output | 164K | 33K |
| Input $ / M tokens | $0.24 | $2 |
| Output $ / M tokens | $0.90 | $8 |
| Results tracked | 60 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
DeepSeek-V3: 42.3 (#106), GPT-4.1: 34.4 (#238)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| Aider Polyglot | 55.1% | 52.4% |
| WeirdML | 36.1% | 39% |
| LMArena Coding | 1368 | 1391 |
| SWE-bench Verified | — | 48.5% |
| SWE-bench Verified (bash only) | — | 39.6% |
| SciCode | 35.8% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| CadEval | — | 42% |
| ALE-Bench | — | 558.1 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, GPT-4.1: 34.7 (#43)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 54% |
| METR Time Horizons | 49.6% | — |
Reasoning DeepSeek-V3 leads
DeepSeek-V3: 20.5 (#236), GPT-4.1: 11.7 (#339)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| SimpleBench | 27.2% | 27% |
| Kagi LLM Benchmark | 52.3% | 52.3% |
| LMArena Hard Prompts | 1365 | 1384 |
| DTBench | 64.8% | 68.3% |
| LMCA | 15.5% | 25.6% |
| Epoch Capabilities Index | 135.94 | 136.78 |
| ForecastBench | 59.1 | 61.5 |
| ARC-AGI-2 | — | 0.4% |
| ARC-AGI-1 | — | 5.5% |
| CritPt | 0% | — |
| Chess Puzzles | — | 6% |
| EnigmaEval | — | 2.2% |
| LiveBench Reasoning | 65.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| BIG-Bench Hard | 87.5% | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math DeepSeek-V3 leads
DeepSeek-V3: 32.1 (#219), GPT-4.1: 22.3 (#280)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 38.3% |
| Omni-MATH | 40.3% | 47.1% |
| LMArena Math | 1373 | 1370 |
| MATH Level 5 | 75.5% | 83% |
| FrontierMath (Feb 2025 set) | 1.7% | 5.5% |
| FrontierMath (Tiers 1-3) | — | 6% |
| LiveBench Math | 73.5% | — |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge Too close to call
DeepSeek-V3: 37.5 (#155), GPT-4.1: 37.1 (#160)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| GPQA Diamond | 67.6% | 66.9% |
| MMLU-Pro | 72.3% | 81.1% |
| Vectara Hallucination Rate | 6.1% | 5.6% |
| GPQA (HELM) | 53.8% | 65.9% |
| LMArena Expert | 1351 | 1364 |
| Humanity's Last Exam | — | 5.4% |
| SimpleQA Verified | — | 31.1% |
| Confabulations | 26.1% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multimodal Not comparable
DeepSeek-V3: —, GPT-4.1: 38.2 (#67)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| LMArena Vision | — | 1211 |
| GeoBench | — | 72% |
Multilingual Too close to call
DeepSeek-V3: 48.5 (#143), GPT-4.1: 49.4 (#133)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| LMArena Non-English | 1358 | 1370 |
| LMArena Chinese | 1391 | 1382 |
| LMArena French | 1385 | 1382 |
| LMArena German | 1374 | 1381 |
| LMArena Japanese | 1333 | 1319 |
| LMArena Korean | 1319 | 1339 |
| LMArena Russian | 1373 | 1377 |
| LMArena Spanish | 1358 | 1376 |
Instruction Following DeepSeek-V3 leads
DeepSeek-V3: 72.8 (#130), GPT-4.1: 71.3 (#153)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| IFEval | 83.2% | 83.8% |
| LMArena Instruction Following | 1345 | 1367 |
| LiveBench Instruction Following | 81.5% | — |
Long Context GPT-4.1 leads
DeepSeek-V3: 34.0 (#253), GPT-4.1: 40.0 (#163)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| Fiction.LiveBench | 50% | 63.9% |
| LMArena Longer Query | 1352 | 1385 |
Writing & Preference Too close to call
DeepSeek-V3: 57.4 (#130), GPT-4.1: 57.6 (#125)
| Benchmark | DeepSeek-V3 | GPT-4.1 |
|---|---|---|
| LMArena Text | 1375 | 1383 |
| LMArena Creative Writing | 1364 | 1363 |
| EQ-Bench Creative Writing | 1472 | 1420 |
| WildBench | 83% | 85.4% |
| LMArena Multi-Turn | 1389 | 1398 |
| Short-Story Creative Writing | 77% | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than GPT-4.1?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 35.9 on the Noometry Index.
Which is cheaper, DeepSeek-V3 or GPT-4.1?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; GPT-4.1 lists at $2 and $8.
Is DeepSeek-V3 or GPT-4.1 better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 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 and GPT-4.1 share?
37 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and GPT-4.1 has 52.