Model comparison
DeepSeek-V3.1 vs GPT-5
GPT-5 is the stronger model overall, scoring 50.9 to 42.8 on the Noometry Index. DeepSeek-V3.1 costs 8.1× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Last verified . 27 shared benchmarks.
DeepSeekSummary
- They share 27 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 2 categories and GPT-5 in 6 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in long context, where GPT-5 leads 69.5 to 36.3.
- The biggest single-benchmark swing is Fiction.LiveBench: 52.8% for DeepSeek-V3.1 and 97.2% for GPT-5.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $1.25 / $10 for GPT-5.
- GPT-5 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 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.8 | 50.9 |
| Released | 2025-08-21 | 2025-08-07 |
| 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 | 69 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 leads
DeepSeek-V3.1: 40.3 (#144), GPT-5: 50.3 (#47)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| WeirdML | 38.4% | 60.7% |
| LMArena Coding | 1417 | 1436 |
| SWE-bench Verified | — | 73.6% |
| SWE-bench Verified (bash only) | — | 65% |
| Aider Polyglot | — | 88% |
| LMArena WebDev | — | 1418 |
| SciCode | — | 42.9% |
| GSO | — | 6.9% |
| ALE-Bench | — | 1,162 |
| AlgoTune | — | 1.67 |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, GPT-5: 33.1 (#56)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| Terminal-Bench | — | 49.6% |
| GDPval | — | 34.8% |
| Remote Labor Index | — | 1.7% |
| DeepResearch Bench | — | 49.6% |
| BALROG | — | 32.8% |
| LMArena Search | — | 1133 |
| METR Time Horizons | — | 69.6% |
Reasoning GPT-5 leads
DeepSeek-V3.1: 27.9 (#110), GPT-5: 38.3 (#64)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| SimpleBench | 40% | 56.7% |
| Kagi LLM Benchmark | 53.2% | 72.7% |
| LMArena Hard Prompts | 1417 | 1416 |
| DTBench | 82.7% | 90.7% |
| LMCA | 24.3% | 40% |
| Epoch Capabilities Index | 139.92 | 150 |
| ForecastBench | 58 | 61.4 |
| ARC-AGI-2 | — | 9.9% |
| ARC-AGI-1 | — | 65.7% |
| CritPt | — | 12.6% |
| Chess Puzzles | — | 37% |
| EnigmaEval | — | 10.5% |
| EBR-Bench | — | 12.7% |
| Mystery Game Puzzles | — | 23% |
Math GPT-5 leads
DeepSeek-V3.1: 38.9 (#122), GPT-5: 55.0 (#44)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| LMArena Math | 1420 | 1407 |
| FrontierMath (Tiers 1-3) | — | 55.4% |
| FrontierMath Tier 4 | — | 22% |
| OTIS Mock AIME 2024-2025 | — | 91.4% |
| ProofBench | — | 18% |
| Omni-MATH | — | 64.7% |
| MATH Level 5 | — | 98.1% |
| FrontierMath (Feb 2025 set) | — | 32.4% |
| FrontierMath Tier 4 (v1) | — | 12.5% |
Knowledge GPT-5 leads
DeepSeek-V3.1: 43.7 (#90), GPT-5: 56.6 (#43)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| Vectara Hallucination Rate | 5.5% | 14.7% |
| LMArena Expert | 1405 | 1419 |
| GPQA Diamond | — | 86.2% |
| Humanity's Last Exam | — | 25.3% |
| SimpleQA Verified | — | 50.1% |
| MMLU-Pro | — | 86.3% |
| Confabulations | — | 10.3% |
| GPQA (HELM) | — | 79.2% |
Multimodal Not comparable
DeepSeek-V3.1: —, GPT-5: 46.8 (#13)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| LMArena Vision | — | 1232 |
| GeoBench | — | 81% |
| VPCT | — | 66% |
Multilingual Too close to call
DeepSeek-V3.1: 51.6 (#106), GPT-5: 51.4 (#110)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| LMArena Non-English | 1400 | 1397 |
| LMArena Chinese | 1469 | 1422 |
| LMArena French | 1447 | 1410 |
| LMArena German | 1411 | 1416 |
| LMArena Japanese | 1378 | 1409 |
| LMArena Korean | 1337 | 1360 |
| LMArena Russian | 1405 | 1406 |
| LMArena Spanish | 1431 | 1399 |
Instruction Following Too close to call
DeepSeek-V3.1: 73.9 (#110), GPT-5: 73.8 (#113)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| LMArena Instruction Following | 1400 | 1388 |
| IFEval | — | 87.5% |
Long Context GPT-5 leads
DeepSeek-V3.1: 36.3 (#232), GPT-5: 69.5 (#2)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| Fiction.LiveBench | 52.8% | 97.2% |
| LMArena Longer Query | 1422 | 1399 |
Writing & Preference GPT-5 leads
DeepSeek-V3.1: 60.3 (#98), GPT-5: 63.4 (#65)
| Benchmark | DeepSeek-V3.1 | GPT-5 |
|---|---|---|
| LMArena Text | 1420 | 1406 |
| LMArena Creative Writing | 1401 | 1365 |
| EQ-Bench Creative Writing | 1436 | 1627 |
| LMArena Multi-Turn | 1408 | 1426 |
| Short-Story Creative Writing | — | 86% |
| WildBench | — | 85.7% |
Frequently asked questions
Is DeepSeek-V3.1 better than GPT-5?
GPT-5 is the stronger model overall, scoring 50.9 to 42.8 on the Noometry Index. DeepSeek-V3.1 costs 8.1× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.1 or GPT-5?
DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; GPT-5 lists at $1.25 and $10.
Is DeepSeek-V3.1 or GPT-5 better for coding?
GPT-5 scores higher on coding benchmarks: 50.3 versus 40.3 in the Noometry coding category.
Which has the bigger context window?
GPT-5 does, with 400K tokens against 164K.
How many benchmarks do DeepSeek-V3.1 and GPT-5 share?
27 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and GPT-5 has 69.