Model comparison
DeepSeek-V3 vs o3-mini
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 36.7 on the Noometry Index.
Last verified . 40 shared benchmarks.
DeepSeekSummary
- They share 40 benchmarks with published results for both. DeepSeek-V3 scores higher in 6 categories and o3-mini in 2 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-V3 leads 57.4 to 50.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 76.9% for o3-mini.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $1.10 / $4.40 for o3-mini.
- o3-mini accepts more context: 200K tokens versus 164K.
- DeepSeek-V3 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3 | o3-mini | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 39.5 | 36.7 |
| Released | 2024-12-26 | 2024-12-20 |
| Weights | Open | Proprietary |
| Context window | 164K | 200K |
| Max output | 164K | 100K |
| Input $ / M tokens | $0.24 | $1.10 |
| Output $ / M tokens | $0.90 | $4.40 |
| Results tracked | 60 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
DeepSeek-V3: 42.3 (#106), o3-mini: 40.8 (#132)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| Aider Polyglot | 55.1% | 60.4% |
| SciCode | 35.8% | 39.8% |
| WeirdML | 36.1% | 43.7% |
| LiveBench Coding | 70.9% | 82.7% |
| LMArena Coding | 1368 | 1378 |
| GSO | — | 1.3% |
| BigCodeBench Instruct | 50% | — |
| BigCodeBench Complete | 62.2% | — |
| CadEval | — | 54% |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, o3-mini: 29.6 (#84)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| Cybench | — | 22.5% |
| METR Time Horizons | 49.6% | — |
Reasoning DeepSeek-V3 leads
DeepSeek-V3: 20.5 (#236), o3-mini: 16.3 (#305)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| SimpleBench | 27.2% | 22.8% |
| CritPt | 0% | 0.3% |
| LiveBench Reasoning | 65.8% | 89.6% |
| LMArena Hard Prompts | 1365 | 1366 |
| DTBench | 64.8% | 68.8% |
| LiveBench Data Analysis | 60.9% | 70.6% |
| LMCA | 15.5% | 19% |
| Epoch Capabilities Index | 135.94 | 140.34 |
| ForecastBench | 59.1 | 59.6 |
| LiveBench | 66.9% | 75.9% |
| ARC-AGI-2 | — | 3% |
| Kagi LLM Benchmark | 52.3% | — |
| ARC-AGI-1 | — | 34.5% |
| Chess Puzzles | — | 17% |
| Mystery Game Puzzles | — | 7% |
| BIG-Bench Hard | 87.5% | — |
| HellaSwag | 88.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math DeepSeek-V3 leads
DeepSeek-V3: 32.1 (#219), o3-mini: 28.1 (#244)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 76.9% |
| LiveBench Math | 73.5% | 77.3% |
| LMArena Math | 1373 | 1396 |
| MATH Level 5 | 75.5% | 96.5% |
| FrontierMath (Feb 2025 set) | 1.7% | 12.4% |
| FrontierMath (Tiers 1-3) | — | 18.6% |
| FrontierMath Tier 4 | — | 0% |
| Omni-MATH | 40.3% | — |
| FrontierMath Tier 4 (v1) | — | 4.2% |
Knowledge Too close to call
DeepSeek-V3: 37.5 (#155), o3-mini: 38.3 (#146)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| GPQA Diamond | 67.6% | 77% |
| Confabulations | 26.1% | 17.9% |
| LMArena Expert | 1351 | 1364 |
| SimpleQA Verified | — | 15.3% |
| MMLU-Pro | 72.3% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multilingual DeepSeek-V3 leads
DeepSeek-V3: 48.5 (#143), o3-mini: 45.7 (#164)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| LMArena Non-English | 1358 | 1319 |
| LMArena Chinese | 1391 | 1379 |
| LMArena French | 1385 | 1334 |
| LMArena German | 1374 | 1303 |
| LMArena Japanese | 1333 | 1286 |
| LMArena Korean | 1319 | 1314 |
| LMArena Russian | 1373 | 1304 |
| LMArena Spanish | 1358 | 1321 |
Instruction Following o3-mini leads
DeepSeek-V3: 72.8 (#130), o3-mini: 75.1 (#72)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| LiveBench Instruction Following | 81.5% | 84.4% |
| LMArena Instruction Following | 1345 | 1337 |
| IFEval | 83.2% | — |
Long Context Too close to call
DeepSeek-V3: 34.0 (#253), o3-mini: 33.8 (#256)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| Fiction.LiveBench | 50% | 50% |
| LMArena Longer Query | 1352 | 1343 |
Writing & Preference DeepSeek-V3 leads
DeepSeek-V3: 57.4 (#130), o3-mini: 50.3 (#182)
| Benchmark | DeepSeek-V3 | o3-mini |
|---|---|---|
| LMArena Text | 1375 | 1337 |
| LMArena Creative Writing | 1364 | 1286 |
| Short-Story Creative Writing | 77% | 61.7% |
| LMArena Multi-Turn | 1389 | 1320 |
| LiveBench Language | 49.1% | 50.7% |
| EQ-Bench Creative Writing | 1472 | — |
| WildBench | 83% | — |
Frequently asked questions
Is DeepSeek-V3 better than o3-mini?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 36.7 on the Noometry Index.
Which is cheaper, DeepSeek-V3 or o3-mini?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; o3-mini lists at $1.10 and $4.40.
Is DeepSeek-V3 or o3-mini better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 versus 40.8 in the Noometry coding category.
Which has the bigger context window?
o3-mini does, with 200K tokens against 164K.
How many benchmarks do DeepSeek-V3 and o3-mini share?
40 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and o3-mini has 51.