Model comparison
DeepSeek-V3.1 vs o3
o3 is the stronger model overall, scoring 47.5 to 42.8 on the Noometry Index. DeepSeek-V3.1 costs 8.2× less per token, which makes it the better buy when o3's lead doesn't matter for your workload.
Last verified . 26 shared benchmarks.
DeepSeekSummary
- They share 26 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 1 category and o3 in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in long context, where o3 leads 53.3 to 36.3.
- The biggest single-benchmark swing is Fiction.LiveBench: 52.8% for DeepSeek-V3.1 and 88.9% for o3.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $2 / $8 for o3.
- o3 accepts more context: 200K tokens versus 164K.
- DeepSeek-V3.1 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.1 | o3 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.8 | 47.5 |
| Released | 2025-08-21 | 2025-04-16 |
| Weights | Open | Proprietary |
| Context window | 164K | 200K |
| Max output | 8K | 100K |
| Input $ / M tokens | $0.25 | $2 |
| Output $ / M tokens | $0.95 | $8 |
| Results tracked | 27 | 63 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o3 leads
DeepSeek-V3.1: 40.3 (#144), o3: 46.8 (#64)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| WeirdML | 38.4% | 52.4% |
| LMArena Coding | 1417 | 1408 |
| SWE-bench Verified | — | 62.3% |
| SWE-bench Verified (bash only) | — | 58.4% |
| Aider Polyglot | — | 81.3% |
| GSO | — | 8.8% |
| CadEval | — | 74% |
| ALE-Bench | — | 933.55 |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, o3: 34.5 (#44)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 30.8% |
| DeepResearch Bench | — | 45.2% |
| OSWorld | — | 23% |
| LMArena Search | — | 1144 |
| METR Time Horizons | — | 65.4% |
Reasoning o3 leads
DeepSeek-V3.1: 27.9 (#110), o3: 32.0 (#78)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| SimpleBench | 40% | 53.1% |
| Kagi LLM Benchmark | 53.2% | 67.6% |
| LMArena Hard Prompts | 1417 | 1402 |
| DTBench | 82.7% | 84.8% |
| LMCA | 24.3% | 39.7% |
| Epoch Capabilities Index | 139.92 | 146.86 |
| ForecastBench | 58 | 62.5 |
| ARC-AGI-2 | — | 6.5% |
| ARC-AGI-1 | — | 60.8% |
| CritPt | — | 1.4% |
| Chess Puzzles | — | 38% |
| EnigmaEval | — | 13.1% |
| Mystery Game Puzzles | — | 29% |
Math o3 leads
DeepSeek-V3.1: 38.9 (#122), o3: 50.2 (#58)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| LMArena Math | 1420 | 1426 |
| FrontierMath (Tiers 1-3) | — | 33.3% |
| OTIS Mock AIME 2024-2025 | — | 84.4% |
| Omni-MATH | — | 71.4% |
| MATH Level 5 | — | 97.8% |
| FrontierMath (Feb 2025 set) | — | 18.7% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge o3 leads
DeepSeek-V3.1: 43.7 (#90), o3: 54.6 (#52)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| LMArena Expert | 1405 | 1402 |
| GPQA Diamond | — | 81.8% |
| Humanity's Last Exam | — | 20.3% |
| SimpleQA Verified | — | 49.4% |
| MMLU-Pro | — | 85.9% |
| Confabulations | — | 14.4% |
| Vectara Hallucination Rate | 5.5% | — |
| GPQA (HELM) | — | 75.3% |
Multimodal Not comparable
DeepSeek-V3.1: —, o3: 41.4 (#36)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| LMArena Vision | — | 1214 |
| GeoBench | — | 74% |
| VPCT | — | 52% |
Multilingual Too close to call
DeepSeek-V3.1: 51.6 (#106), o3: 51.7 (#105)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| LMArena Non-English | 1400 | 1401 |
| LMArena Chinese | 1469 | 1437 |
| LMArena French | 1447 | 1430 |
| LMArena German | 1411 | 1420 |
| LMArena Japanese | 1378 | 1403 |
| LMArena Korean | 1337 | 1370 |
| LMArena Russian | 1405 | 1406 |
| LMArena Spanish | 1431 | 1395 |
Instruction Following DeepSeek-V3.1 leads
DeepSeek-V3.1: 73.9 (#110), o3: 72.8 (#127)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| LMArena Instruction Following | 1400 | 1368 |
| IFEval | — | 86.9% |
Long Context o3 leads
DeepSeek-V3.1: 36.3 (#232), o3: 53.3 (#6)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| Fiction.LiveBench | 52.8% | 88.9% |
| LMArena Longer Query | 1422 | 1372 |
| CL-bench | — | 17.8% |
Writing & Preference o3 leads
DeepSeek-V3.1: 60.3 (#98), o3: 63.5 (#64)
| Benchmark | DeepSeek-V3.1 | o3 |
|---|---|---|
| LMArena Text | 1420 | 1410 |
| LMArena Creative Writing | 1401 | 1359 |
| EQ-Bench Creative Writing | 1436 | 1676 |
| LMArena Multi-Turn | 1408 | 1405 |
| Short-Story Creative Writing | — | 83.9% |
| WildBench | — | 86.1% |
Frequently asked questions
Is DeepSeek-V3.1 better than o3?
o3 is the stronger model overall, scoring 47.5 to 42.8 on the Noometry Index. DeepSeek-V3.1 costs 8.2× less per token, which makes it the better buy when o3's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.1 or o3?
DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; o3 lists at $2 and $8.
Is DeepSeek-V3.1 or o3 better for coding?
o3 scores higher on coding benchmarks: 46.8 versus 40.3 in the Noometry coding category.
Which has the bigger context window?
o3 does, with 200K tokens against 164K.
How many benchmarks do DeepSeek-V3.1 and o3 share?
26 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and o3 has 63.