Model comparison
DeepSeek-V3 vs GPT-5 Nano
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 33.5 on the Noometry Index. GPT-5 Nano costs 2.9× less per token, which makes it the better buy when DeepSeek-V3's lead doesn't matter for your workload.
Last verified . 34 shared benchmarks.
DeepSeekSummary
- They share 34 benchmarks with published results for both. DeepSeek-V3 scores higher in 7 categories and GPT-5 Nano in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-V3 leads 57.4 to 39.1.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 81.1% for GPT-5 Nano.
- GPT-5 Nano is cheaper at $0.05 / $0.40 per million input/output tokens, against $0.24 / $0.90 for DeepSeek-V3.
- GPT-5 Nano accepts more context: 400K tokens versus 164K.
- DeepSeek-V3 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3 | GPT-5 Nano | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 39.5 | 33.5 |
| Released | 2024-12-26 | 2025-08-07 |
| Weights | Open | Proprietary |
| Context window | 164K | 400K |
| Max output | 164K | 128K |
| Input $ / M tokens | $0.24 | $0.05 |
| Output $ / M tokens | $0.90 | $0.40 |
| Results tracked | 60 | 49 |
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-5 Nano: 33.6 (#254)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| WeirdML | 36.1% | 38.1% |
| LMArena Coding | 1368 | 1351 |
| SWE-bench Verified (bash only) | — | 34.8% |
| Aider Polyglot | 55.1% | — |
| SciCode | 35.8% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 718.67 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, GPT-5 Nano: 25.8 (#106)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| Terminal-Bench | — | 21.8% |
| Berkeley Function Calling Leaderboard | — | 51.5% |
| METR Time Horizons | 49.6% | — |
Reasoning DeepSeek-V3 leads
DeepSeek-V3: 20.5 (#236), GPT-5 Nano: 16.3 (#306)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 62.2% |
| LMArena Hard Prompts | 1365 | 1328 |
| DTBench | 64.8% | 62.7% |
| LMCA | 15.5% | 7.9% |
| Epoch Capabilities Index | 135.94 | 139.38 |
| ForecastBench | 59.1 | 59.1 |
| ARC-AGI-2 | — | 2.6% |
| SimpleBench | 27.2% | — |
| ARC-AGI-1 | — | 20.7% |
| CritPt | 0% | — |
| Chess Puzzles | — | 27% |
| LiveBench Reasoning | 65.8% | — |
| Mystery Game Puzzles | — | 9% |
| 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-5 Nano: 29.4 (#241)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 81.1% |
| Omni-MATH | 40.3% | 54.6% |
| LMArena Math | 1373 | 1317 |
| MATH Level 5 | 75.5% | 95.2% |
| FrontierMath (Feb 2025 set) | 1.7% | 8.3% |
| FrontierMath (Tiers 1-3) | — | 20% |
| FrontierMath Tier 4 | — | 2.4% |
| ProofBench | — | 12% |
| LiveBench Math | 73.5% | — |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge DeepSeek-V3 leads
DeepSeek-V3: 37.5 (#155), GPT-5 Nano: 35.9 (#178)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| GPQA Diamond | 67.6% | 69.4% |
| MMLU-Pro | 72.3% | 77.8% |
| Vectara Hallucination Rate | 6.1% | 10.5% |
| GPQA (HELM) | 53.8% | 67.9% |
| LMArena Expert | 1351 | 1321 |
| SimpleQA Verified | — | 11.7% |
| Confabulations | 26.1% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multimodal Not comparable
DeepSeek-V3: —, GPT-5 Nano: 31.3 (#108)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| LMArena Vision | — | 1159 |
| VPCT | — | 37.2% |
Multilingual DeepSeek-V3 leads
DeepSeek-V3: 48.5 (#143), GPT-5 Nano: 45.3 (#172)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| LMArena Non-English | 1358 | 1313 |
| LMArena Chinese | 1391 | 1356 |
| LMArena German | 1374 | 1327 |
| LMArena Japanese | 1333 | 1226 |
| LMArena Korean | 1319 | 1269 |
| LMArena Russian | 1373 | 1296 |
| LMArena Spanish | 1358 | 1360 |
| LMArena French | 1385 | — |
Instruction Following GPT-5 Nano leads
DeepSeek-V3: 72.8 (#130), GPT-5 Nano: 75.0 (#79)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| IFEval | 83.2% | 93.2% |
| LMArena Instruction Following | 1345 | 1306 |
| LiveBench Instruction Following | 81.5% | — |
Long Context DeepSeek-V3 leads
DeepSeek-V3: 34.0 (#253), GPT-5 Nano: 31.3 (#281)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| Fiction.LiveBench | 50% | 44.4% |
| LMArena Longer Query | 1352 | 1312 |
Writing & Preference DeepSeek-V3 leads
DeepSeek-V3: 57.4 (#130), GPT-5 Nano: 39.1 (#249)
| Benchmark | DeepSeek-V3 | GPT-5 Nano |
|---|---|---|
| LMArena Text | 1375 | 1320 |
| LMArena Creative Writing | 1364 | 1249 |
| EQ-Bench Creative Writing | 1472 | 705 |
| WildBench | 83% | 80.6% |
| LMArena Multi-Turn | 1389 | 1311 |
| Short-Story Creative Writing | 77% | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than GPT-5 Nano?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 33.5 on the Noometry Index. GPT-5 Nano costs 2.9× less per token, which makes it the better buy when DeepSeek-V3's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3 or GPT-5 Nano?
GPT-5 Nano is cheaper. It lists at $0.05 per million input tokens and $0.40 per million output tokens; DeepSeek-V3 lists at $0.24 and $0.90.
Is DeepSeek-V3 or GPT-5 Nano better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 versus 33.6 in the Noometry coding category.
Which has the bigger context window?
GPT-5 Nano does, with 400K tokens against 164K.
How many benchmarks do DeepSeek-V3 and GPT-5 Nano share?
34 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and GPT-5 Nano has 49.