# Gemini 3 Flash Preview vs gpt-oss-20b

> Gemini 3 Flash Preview is the stronger model overall, scoring 52.3 to 32.1 on the Noometry Index. gpt-oss-20b costs 31× less per token, which makes it the better buy when Gemini 3 Flash Preview's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gemini-3-flash-preview-vs-gpt-oss-20b
- Last updated: 2026-10-10
- Shared benchmarks: 25

## Summary

- They share 25 benchmarks with published results for both. Gemini 3 Flash Preview scores higher in 9 categories and gpt-oss-20b in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Gemini 3 Flash Preview leads 49.2 to 18.8.
- The biggest single-benchmark swing is Terminal-Bench: 64.3% for Gemini 3 Flash Preview and 3.4% for gpt-oss-20b.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $0.50 / $3 for Gemini 3 Flash Preview.
- Gemini 3 Flash Preview accepts more context: 1.05M tokens versus 131K.
- gpt-oss-20b has downloadable open weights; the other is API-only.

## Snapshot

| | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| Provider | Google | OpenAI |
| Noometry Index | 52.3 | 32.1 |
| Rank | 33 | 249 |
| Context | 1.05M | 131K |
| Input $/M | $0.50 | $0.018 |
| Output $/M | $3 | $0.09 |
| Weights | Proprietary | Open |

## Coding

- Gemini 3 Flash Preview: 51.5 (#35)
- gpt-oss-20b: 37.1 (#185)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| WeirdML | 61.6% | 40.9% |
| LMArena Coding | 1460 | 1307 |
| ALE-Bench | 1,367 | 566.05 |
| SWE-bench Verified | 75.4% | — |
| SWE-bench Verified (bash only) | 75.8% | — |
| LMArena WebDev | 1439 | — |
| SWE-bench Multilingual | 72.7% | — |
| SciCode | — | 34.4% |
| GSO | 9.8% | — |

## Agentic & Tool Use

- Gemini 3 Flash Preview: 38.6 (#27)
- gpt-oss-20b: 9.2 (#146)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| Terminal-Bench | 64.3% | 3.4% |
| τ²-bench Airline | 82.5% | — |
| τ²-bench Banking | 27.3% | — |
| τ²-bench Retail | 76.8% | — |
| τ²-bench Telecom | 91.2% | — |
| DeepResearch Bench | 49.8% | — |
| BALROG | 48.1% | — |
| GDP.pdf | 10% | — |
| LMArena Search | 1198 | — |
| Vending-Bench 2 | 3,635 | — |

## Reasoning

- Gemini 3 Flash Preview: 49.2 (#35)
- gpt-oss-20b: 18.8 (#251)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| Chess Puzzles | 40% | 4% |
| LMArena Hard Prompts | 1465 | 1274 |
| DTBench | 89.1% | 68% |
| LMCA | 43.1% | 14.5% |
| Epoch Capabilities Index | 151.83 | 137.81 |
| ARC-AGI-2 | 33.6% | — |
| SimpleBench | 61.1% | — |
| Kagi LLM Benchmark | — | 53.2% |
| NYT Connections (extended) | 83.1% | — |
| ARC-AGI-1 | 84.7% | — |
| CritPt | — | 1.4% |
| Mystery Game Puzzles | 26% | — |
| ForecastBench | 58.5 | — |

## Math

- Gemini 3 Flash Preview: 51.4 (#42)
- gpt-oss-20b: 37.6 (#93)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 95.6% | 65.3% |
| LMArena Math | 1472 | 1317 |
| FrontierMath (Tiers 1-3) | 51.2% | — |
| FrontierMath Tier 4 | 17.1% | — |
| MathArena Final-Answer Competitions | 67.6% | — |
| ProofBench | 15% | — |
| Omni-MATH | — | 56.5% |
| FrontierMath (Feb 2025 set) | 35.6% | — |
| FrontierMath Tier 4 (v1) | 4.2% | — |

## Knowledge

- Gemini 3 Flash Preview: 58.8 (#30)
- gpt-oss-20b: 34.6 (#186)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| GPQA Diamond | 89.4% | 60.8% |
| LMArena Expert | 1461 | 1258 |
| SimpleQA Verified | 66.8% | — |
| MMLU-Pro | — | 74% |
| Vectara Hallucination Rate | 13.5% | — |
| GPQA (HELM) | — | 59.4% |

## Multimodal

- Gemini 3 Flash Preview: 45.3 (#15)
- gpt-oss-20b: —

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| LMArena Vision | 1285 | — |
| GeoBench | 88% | — |
| VPCT | 72.6% | — |
| Blueprint-Bench 2 | 0% | — |
| LMArena Document | 1413 | — |

## Multilingual

- Gemini 3 Flash Preview: 55.9 (#26)
- gpt-oss-20b: 42.4 (#192)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| LMArena Non-English | 1458 | 1269 |
| LMArena Chinese | 1510 | 1311 |
| LMArena German | 1497 | 1249 |
| LMArena Japanese | 1488 | 1245 |
| LMArena Korean | 1443 | 1239 |
| LMArena Russian | 1480 | 1279 |
| LMArena Spanish | 1470 | 1267 |
| LMArena French | 1478 | — |

## Instruction Following

- Gemini 3 Flash Preview: 75.8 (#52)
- gpt-oss-20b: 62.0 (#235)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| LMArena Instruction Following | 1437 | 1236 |
| IFEval | — | 73.2% |

## Long Context

- Gemini 3 Flash Preview: 44.4 (#63)
- gpt-oss-20b: 37.9 (#203)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| LMArena Longer Query | 1452 | 1251 |

## Writing & Preference

- Gemini 3 Flash Preview: 65.6 (#44)
- gpt-oss-20b: 35.8 (#260)

| Benchmark | Gemini 3 Flash Preview | gpt-oss-20b |
|---|---|---|
| LMArena Text | 1466 | 1288 |
| LMArena Creative Writing | 1457 | 1201 |
| LMArena Multi-Turn | 1471 | 1268 |
| EQ-Bench Creative Writing | — | 666 |
| WildBench | — | 73.7% |

## FAQ

### Is Gemini 3 Flash Preview better than gpt-oss-20b?

Gemini 3 Flash Preview is the stronger model overall, scoring 52.3 to 32.1 on the Noometry Index. gpt-oss-20b costs 31× less per token, which makes it the better buy when Gemini 3 Flash Preview's lead doesn't matter for your workload.

### Which is cheaper, Gemini 3 Flash Preview 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; Gemini 3 Flash Preview lists at $0.50 and $3.

### Is Gemini 3 Flash Preview or gpt-oss-20b better for coding?

Gemini 3 Flash Preview scores higher on coding benchmarks: 51.5 versus 37.1 in the Noometry coding category.

### Which has the bigger context window?

Gemini 3 Flash Preview does, with 1.05M tokens against 131K.

### How many benchmarks do Gemini 3 Flash Preview and gpt-oss-20b share?

25 benchmarks have published results for both models. Gemini 3 Flash Preview has 59 scored results on Noometry and gpt-oss-20b has 34.