Model comparison
DeepSeek-V3.1 vs GPT-5.1
GPT-5.1 is the stronger model overall, scoring 48.8 to 42.2 on the Noometry Index. DeepSeek-V3.1 costs 8.1× less per token, which makes it the better buy when GPT-5.1's lead doesn't matter for your workload.
Last verified . 24 shared benchmarks.
DeepSeekSummary
- They share 24 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 0 categories and GPT-5.1 in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.1 leads 50.3 to 36.1.
- The biggest single-benchmark swing is WeirdML: 38.4% for DeepSeek-V3.1 and 60.8% for GPT-5.1.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $1.25 / $10 for GPT-5.1.
- GPT-5.1 accepts more context: 400K tokens versus 164K.
- DeepSeek-V3.1 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.1 | GPT-5.1 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.2 | 48.8 |
| Released | 2025-08-21 | 2025-11-13 |
| Weights | Open | Proprietary |
| Context window | 164K | 400K |
| Max output | 8K | 128K |
| Input $ / M tokens | $0.25 | $1.25 |
| Output $ / M tokens | $0.95 | $10 |
| Results tracked | 27 | 63 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.1 leads
DeepSeek-V3.1: 39.8 (#140), GPT-5.1: 46.9 (#59)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| WeirdML | 38.4% | 60.8% |
| LMArena Coding | 1417 | 1454 |
| SWE-bench Verified | — | 68% |
| SWE-bench Verified (bash only) | — | 66% |
| LMArena WebDev | — | 1395 |
| SciCode | — | 43.3% |
| GSO | — | 13.7% |
| LiveBench Coding | — | 72.5% |
| ALE-Bench | — | 1,192 |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, GPT-5.1: 32.5 (#55)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| Terminal-Bench | — | 47.6% |
| DeepResearch Bench | — | 42.8% |
| LMArena Search | — | 1199 |
| Vending-Bench 2 | — | 1,473 |
Reasoning GPT-5.1 leads
DeepSeek-V3.1: 27.3 (#109), GPT-5.1: 39.9 (#50)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| SimpleBench | 40% | 53.2% |
| LMArena Hard Prompts | 1417 | 1457 |
| DTBench | 82.7% | 90.1% |
| LMCA | 24.3% | 43.9% |
| Epoch Capabilities Index | 139.92 | 149.66 |
| ForecastBench | 58 | 58.1 |
| ARC-AGI-2 | — | 17.6% |
| Kagi LLM Benchmark | 53.2% | — |
| ARC-AGI-1 | — | 72.8% |
| CritPt | — | 4.9% |
| Chess Puzzles | — | 32% |
| EnigmaEval | — | 11.2% |
| LiveBench Reasoning | — | 95.8% |
| Mystery Game Puzzles | — | 19% |
| LiveBench Data Analysis | — | 72.1% |
| LiveBench | — | 78.8% |
Math GPT-5.1 leads
DeepSeek-V3.1: 36.1 (#125), GPT-5.1: 50.3 (#45)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| LMArena Math | 1420 | 1446 |
| OTIS Mock AIME 2024-2025 | — | 88.6% |
| Omni-MATH | — | 46.4% |
| LiveBench Math | — | 94.5% |
| FrontierMath (Feb 2025 set) | — | 31% |
| FrontierMath Tier 4 (v1) | — | 12.5% |
Knowledge GPT-5.1 leads
DeepSeek-V3.1: 43.0 (#86), GPT-5.1: 50.5 (#65)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| Vectara Hallucination Rate | 5.5% | 10.9% |
| LMArena Expert | 1403 | 1470 |
| GPQA Diamond | — | 87.6% |
| Humanity's Last Exam | — | 23.7% |
| SimpleQA Verified | — | 48% |
| MMLU-Pro | — | 57.9% |
| GPQA (HELM) | — | 44.2% |
Multimodal Not comparable
DeepSeek-V3.1: —, GPT-5.1: 44.0 (#17)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| LMArena Vision | — | 1250 |
| VPCT | — | 58.7% |
| LMArena Document | — | 1403 |
Multilingual GPT-5.1 leads
DeepSeek-V3.1: 51.8 (#100), GPT-5.1: 54.0 (#51)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| LMArena Non-English | 1400 | 1431 |
| LMArena Chinese | 1470 | 1495 |
| LMArena French | 1447 | 1451 |
| LMArena German | 1412 | 1439 |
| LMArena Japanese | 1374 | 1453 |
| LMArena Korean | 1337 | 1401 |
| LMArena Russian | 1405 | 1435 |
| LMArena Spanish | 1429 | 1430 |
Instruction Following GPT-5.1 leads
DeepSeek-V3.1: 74.1 (#105), GPT-5.1: 84.0 (#1)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| LMArena Instruction Following | 1400 | 1442 |
| LiveBench Instruction Following | — | 93.3% |
| IFEval | — | 93.5% |
Long Context GPT-5.1 leads
DeepSeek-V3.1: 36.3 (#227), GPT-5.1: 47.6 (#14)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| LMArena Longer Query | 1422 | 1448 |
| Fiction.LiveBench | 52.8% | — |
| CL-bench | — | 23.7% |
| CL-bench Life | — | 17.3% |
Writing & Preference GPT-5.1 leads
DeepSeek-V3.1: 60.5 (#93), GPT-5.1: 64.7 (#53)
| Benchmark | DeepSeek-V3.1 | GPT-5.1 |
|---|---|---|
| LMArena Text | 1420 | 1443 |
| LMArena Creative Writing | 1401 | 1427 |
| LMArena Multi-Turn | 1408 | 1449 |
| EQ-Bench Creative Writing | 1436 | — |
| WildBench | — | 86.3% |
| LiveBench Language | — | 80.2% |
Frequently asked questions
Is DeepSeek-V3.1 better than GPT-5.1?
GPT-5.1 is the stronger model overall, scoring 48.8 to 42.2 on the Noometry Index. DeepSeek-V3.1 costs 8.1× less per token, which makes it the better buy when GPT-5.1's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.1 or GPT-5.1?
DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; GPT-5.1 lists at $1.25 and $10.
Is DeepSeek-V3.1 or GPT-5.1 better for coding?
GPT-5.1 scores higher on coding benchmarks: 46.9 versus 39.8 in the Noometry coding category.
Which has the bigger context window?
GPT-5.1 does, with 400K tokens against 164K.
How many benchmarks do DeepSeek-V3.1 and GPT-5.1 share?
24 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and GPT-5.1 has 63.