Model comparison
DeepSeek-V3 vs o1
o1 is the stronger model overall, scoring 40.9 to 39.5 on the Noometry Index. DeepSeek-V3 costs 65× less per token, which makes it the better buy when o1's lead doesn't matter for your workload.
Last verified . 40 shared benchmarks.
DeepSeekSummary
- They share 40 benchmarks with published results for both. DeepSeek-V3 scores higher in 1 category and o1 in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in long context, where o1 leads 50.3 to 34.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 73.3% for o1.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $15 / $60 for o1.
- o1 accepts more context: 200K tokens versus 164K.
- DeepSeek-V3 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3 | o1 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 39.5 | 40.9 |
| Released | 2024-12-26 | 2024-09-12 |
| Weights | Open | Proprietary |
| Context window | 164K | 200K |
| Max output | 164K | 100K |
| Input $ / M tokens | $0.24 | $15 |
| Output $ / M tokens | $0.90 | $60 |
| Results tracked | 60 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o1 leads
DeepSeek-V3: 42.3 (#106), o1: 46.1 (#70)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| Aider Polyglot | 55.1% | 61.7% |
| WeirdML | 36.1% | 47.6% |
| LiveBench Coding | 70.9% | 69.7% |
| LMArena Coding | 1368 | 1367 |
| HumanEval+ | 86.6% | 89% |
| MBPP+ | 73% | 80.2% |
| SciCode | 35.8% | — |
| BigCodeBench Instruct | 50% | — |
| BigCodeBench Complete | 62.2% | — |
| CadEval | — | 56% |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, o1: 24.6 (#117)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| METR Time Horizons | 49.6% | 51.1% |
| Cybench | — | 10% |
Reasoning o1 leads
DeepSeek-V3: 20.5 (#236), o1: 27.9 (#111)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| SimpleBench | 27.2% | 41.7% |
| LiveBench Reasoning | 65.8% | 91.6% |
| LMArena Hard Prompts | 1365 | 1371 |
| DTBench | 64.8% | 74.7% |
| LiveBench Data Analysis | 60.9% | 65.5% |
| LMCA | 15.5% | 22.3% |
| Epoch Capabilities Index | 135.94 | 141.91 |
| LiveBench | 66.9% | 75.7% |
| Kagi LLM Benchmark | 52.3% | — |
| ARC-AGI-1 | — | 30.7% |
| CritPt | 0% | — |
| Chess Puzzles | — | 15% |
| EnigmaEval | — | 5.7% |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math o1 leads
DeepSeek-V3: 32.1 (#219), o1: 36.1 (#175)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 73.3% |
| LiveBench Math | 73.5% | 80.3% |
| LMArena Math | 1373 | 1388 |
| MATH Level 5 | 75.5% | 94.7% |
| FrontierMath (Feb 2025 set) | 1.7% | 9.3% |
| FrontierMath (Tiers 1-3) | — | 14.7% |
| Omni-MATH | 40.3% | — |
Knowledge o1 leads
DeepSeek-V3: 37.5 (#155), o1: 41.5 (#110)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| GPQA Diamond | 67.6% | 76.8% |
| Confabulations | 26.1% | 11.7% |
| LMArena Expert | 1351 | 1361 |
| Humanity's Last Exam | — | 8% |
| SimpleQA Verified | — | 41.1% |
| MMLU-Pro | 72.3% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multimodal Not comparable
DeepSeek-V3: —, o1: 34.2 (#93)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| LMArena Vision | — | 1168 |
| GeoBench | — | 80% |
| VPCT | — | 37% |
| SpatialViz-Bench | — | 41.4% |
Multilingual Too close to call
DeepSeek-V3: 48.5 (#143), o1: 48.6 (#142)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| LMArena Non-English | 1358 | 1358 |
| LMArena Chinese | 1391 | 1394 |
| LMArena French | 1385 | 1344 |
| LMArena German | 1374 | 1337 |
| LMArena Japanese | 1333 | 1346 |
| LMArena Korean | 1319 | 1396 |
| LMArena Russian | 1373 | 1356 |
| LMArena Spanish | 1358 | 1345 |
Instruction Following o1 leads
DeepSeek-V3: 72.8 (#130), o1: 74.8 (#86)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| LiveBench Instruction Following | 81.5% | 81.5% |
| LMArena Instruction Following | 1345 | 1367 |
| IFEval | 83.2% | — |
Long Context o1 leads
DeepSeek-V3: 34.0 (#253), o1: 50.3 (#9)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| Fiction.LiveBench | 50% | 83.3% |
| LMArena Longer Query | 1352 | 1378 |
Writing & Preference DeepSeek-V3 leads
DeepSeek-V3: 57.4 (#130), o1: 55.6 (#144)
| Benchmark | DeepSeek-V3 | o1 |
|---|---|---|
| LMArena Text | 1375 | 1366 |
| LMArena Creative Writing | 1364 | 1348 |
| Short-Story Creative Writing | 77% | 70.2% |
| LMArena Multi-Turn | 1389 | 1369 |
| LiveBench Language | 49.1% | 65.4% |
| EQ-Bench Creative Writing | 1472 | — |
| WildBench | 83% | — |
Frequently asked questions
Is DeepSeek-V3 better than o1?
o1 is the stronger model overall, scoring 40.9 to 39.5 on the Noometry Index. DeepSeek-V3 costs 65× less per token, which makes it the better buy when o1's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3 or o1?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; o1 lists at $15 and $60.
Is DeepSeek-V3 or o1 better for coding?
o1 scores higher on coding benchmarks: 46.1 versus 42.3 in the Noometry coding category.
Which has the bigger context window?
o1 does, with 200K tokens against 164K.
How many benchmarks do DeepSeek-V3 and o1 share?
40 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and o1 has 52.