# DeepSeek-V3 vs gpt-oss-20b

> DeepSeek-V3 is the stronger model overall, scoring 39.2 to 32.1 on the Noometry Index. gpt-oss-20b costs 11× less per token, which makes it the better buy when DeepSeek-V3's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/deepseek-v3-vs-gpt-oss-20b
- Last updated: 2026-10-10
- Shared benchmarks: 31

## Summary

- They share 31 benchmarks with published results for both. DeepSeek-V3 scores higher in 6 categories and gpt-oss-20b in 2 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-V3 leads 57.6 to 35.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 65.3% for gpt-oss-20b.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $0.24 / $0.90 for DeepSeek-V3.
- DeepSeek-V3 accepts more context: 164K tokens versus 131K.

## Snapshot

| | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 39.2 | 32.1 |
| Rank | 157 | 249 |
| Context | 164K | 131K |
| Input $/M | $0.24 | $0.018 |
| Output $/M | $0.90 | $0.09 |
| Weights | Open | Open |

## Coding

- DeepSeek-V3: 41.9 (#99)
- gpt-oss-20b: 37.1 (#185)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| SciCode | 35.8% | 34.4% |
| WeirdML | 36.1% | 40.9% |
| LMArena Coding | 1368 | 1307 |
| Aider Polyglot | 55.1% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 566.05 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |

## Agentic & Tool Use

- DeepSeek-V3: —
- gpt-oss-20b: 9.2 (#146)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| Terminal-Bench | — | 3.4% |
| METR Time Horizons | 49.6% | — |

## Reasoning

- DeepSeek-V3: 20.0 (#226)
- gpt-oss-20b: 18.8 (#251)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 53.2% |
| CritPt | 0% | 1.4% |
| LMArena Hard Prompts | 1365 | 1274 |
| DTBench | 64.8% | 68% |
| LMCA | 15.5% | 14.5% |
| Epoch Capabilities Index | 135.95 | 137.81 |
| SimpleBench | 27.2% | — |
| Chess Puzzles | — | 4% |
| LiveBench Reasoning | 65.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |

## Math

- DeepSeek-V3: 30.4 (#208)
- gpt-oss-20b: 37.6 (#93)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 65.3% |
| Omni-MATH | 40.3% | 56.5% |
| LMArena Math | 1373 | 1317 |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |

## Knowledge

- DeepSeek-V3: 37.5 (#147)
- gpt-oss-20b: 34.6 (#186)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| GPQA Diamond | 67.6% | 60.8% |
| MMLU-Pro | 72.3% | 74% |
| GPQA (HELM) | 53.8% | 59.4% |
| LMArena Expert | 1352 | 1258 |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |

## Multilingual

- DeepSeek-V3: 48.8 (#138)
- gpt-oss-20b: 42.4 (#192)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| LMArena Non-English | 1358 | 1269 |
| LMArena Chinese | 1390 | 1311 |
| LMArena German | 1374 | 1249 |
| LMArena Japanese | 1333 | 1245 |
| LMArena Korean | 1321 | 1239 |
| LMArena Russian | 1372 | 1279 |
| LMArena Spanish | 1358 | 1267 |
| LMArena French | 1386 | — |

## Instruction Following

- DeepSeek-V3: 73.0 (#125)
- gpt-oss-20b: 62.0 (#235)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| IFEval | 83.2% | 73.2% |
| LMArena Instruction Following | 1345 | 1236 |
| LiveBench Instruction Following | 81.5% | — |

## Long Context

- DeepSeek-V3: 34.0 (#248)
- gpt-oss-20b: 37.9 (#203)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| LMArena Longer Query | 1353 | 1251 |
| Fiction.LiveBench | 50% | — |

## Writing & Preference

- DeepSeek-V3: 57.6 (#124)
- gpt-oss-20b: 35.8 (#260)

| Benchmark | DeepSeek-V3 | gpt-oss-20b |
|---|---|---|
| LMArena Text | 1375 | 1288 |
| LMArena Creative Writing | 1364 | 1201 |
| EQ-Bench Creative Writing | 1472 | 666 |
| WildBench | 83% | 73.7% |
| LMArena Multi-Turn | 1389 | 1268 |
| Short-Story Creative Writing | 77% | — |
| LiveBench Language | 49.1% | — |

## FAQ

### Is DeepSeek-V3 better than gpt-oss-20b?

DeepSeek-V3 is the stronger model overall, scoring 39.2 to 32.1 on the Noometry Index. gpt-oss-20b costs 11× 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-oss-20b?

gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; DeepSeek-V3 lists at $0.24 and $0.90.

### Is DeepSeek-V3 or gpt-oss-20b better for coding?

DeepSeek-V3 scores higher on coding benchmarks: 41.9 versus 37.1 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek-V3 does, with 164K tokens against 131K.

### How many benchmarks do DeepSeek-V3 and gpt-oss-20b share?

31 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and gpt-oss-20b has 34.