# gpt-oss-20b vs MiniMax-M3

> MiniMax-M3 is the stronger model overall, scoring 43.5 to 32.1 on the Noometry Index. gpt-oss-20b costs 15× less per token, which makes it the better buy when MiniMax-M3's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-oss-20b-vs-minimax-m3
- Last updated: 2026-10-10
- Shared benchmarks: 25

## Summary

- They share 25 benchmarks with published results for both. gpt-oss-20b scores higher in 0 categories and MiniMax-M3 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where MiniMax-M3 leads 62.3 to 35.8.
- The biggest single-benchmark swing is GPQA Diamond: 60.8% for gpt-oss-20b and 90.9% for MiniMax-M3.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $0.30 / $1.20 for MiniMax-M3.
- MiniMax-M3 accepts more context: 1M tokens versus 131K.

## Snapshot

| | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| Provider | OpenAI | MiniMax |
| Noometry Index | 32.1 | 43.5 |
| Rank | 249 | 81 |
| Context | 131K | 1M |
| Input $/M | $0.018 | $0.30 |
| Output $/M | $0.09 | $1.20 |
| Weights | Open | Open |

## Coding

- gpt-oss-20b: 37.1 (#185)
- MiniMax-M3: 41.5 (#111)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| SciCode | 34.4% | 47.1% |
| LMArena Coding | 1307 | 1470 |
| ALE-Bench | 566.05 | 640.02 |
| FrontierCode | — | 14.7% |
| LMArena WebDev | — | 1482 |
| WeirdML | 40.9% | — |

## Agentic & Tool Use

- gpt-oss-20b: 9.2 (#146)
- MiniMax-M3: 22.5 (#121)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| Terminal-Bench | 3.4% | — |
| APEX-Agents | — | 37.7% |
| OSWorld 2.0 | — | 4.6% |
| GBAEval | — | 0.9% |
| Vending-Bench 2 | — | 2,158 |

## Reasoning

- gpt-oss-20b: 18.8 (#251)
- MiniMax-M3: 29.3 (#83)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| CritPt | 1.4% | 3.7% |
| Chess Puzzles | 4% | 14% |
| LMArena Hard Prompts | 1274 | 1447 |
| DTBench | 68% | 78.9% |
| LMCA | 14.5% | 33.7% |
| Epoch Capabilities Index | 137.81 | 147 |
| SimpleBench | — | 45.8% |
| Kagi LLM Benchmark | 53.2% | — |
| NYT Connections (extended) | — | 65.1% |
| Mystery Game Puzzles | — | 8% |
| Surface Evolver Bench | — | 55% |
| ForecastBench | — | 61.4 |

## Math

- gpt-oss-20b: 37.6 (#93)
- MiniMax-M3: 39.1 (#83)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 65.3% | 71.1% |
| LMArena Math | 1317 | 1432 |
| ProofBench | — | 18% |
| Omni-MATH | 56.5% | — |

## Knowledge

- gpt-oss-20b: 34.6 (#186)
- MiniMax-M3: 58.4 (#32)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| GPQA Diamond | 60.8% | 90.9% |
| LMArena Expert | 1258 | 1462 |
| MMLU-Pro | 74% | — |
| GPQA (HELM) | 59.4% | — |

## Multimodal

- gpt-oss-20b: —
- MiniMax-M3: 39.5 (#45)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| LMArena Vision | — | 1254 |
| LMArena Document | — | 1435 |

## Multilingual

- gpt-oss-20b: 42.4 (#192)
- MiniMax-M3: 53.2 (#70)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| LMArena Non-English | 1269 | 1420 |
| LMArena Chinese | 1311 | 1465 |
| LMArena German | 1249 | 1424 |
| LMArena Japanese | 1245 | 1381 |
| LMArena Korean | 1239 | 1372 |
| LMArena Russian | 1279 | 1430 |
| LMArena Spanish | 1267 | 1431 |
| LMArena French | — | 1445 |

## Instruction Following

- gpt-oss-20b: 62.0 (#235)
- MiniMax-M3: 75.7 (#56)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| LMArena Instruction Following | 1236 | 1433 |
| IFEval | 73.2% | — |

## Long Context

- gpt-oss-20b: 37.9 (#203)
- MiniMax-M3: 44.2 (#68)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| LMArena Longer Query | 1251 | 1445 |

## Writing & Preference

- gpt-oss-20b: 35.8 (#260)
- MiniMax-M3: 62.3 (#79)

| Benchmark | gpt-oss-20b | MiniMax-M3 |
|---|---|---|
| LMArena Text | 1288 | 1433 |
| LMArena Creative Writing | 1201 | 1405 |
| LMArena Multi-Turn | 1268 | 1441 |
| EQ-Bench Creative Writing | 666 | — |
| WildBench | 73.7% | — |
| EQ-Bench 4 | — | 1150 |

## FAQ

### Is gpt-oss-20b better than MiniMax-M3?

MiniMax-M3 is the stronger model overall, scoring 43.5 to 32.1 on the Noometry Index. gpt-oss-20b costs 15× less per token, which makes it the better buy when MiniMax-M3's lead doesn't matter for your workload.

### Which is cheaper, gpt-oss-20b or MiniMax-M3?

gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; MiniMax-M3 lists at $0.30 and $1.20.

### Is gpt-oss-20b or MiniMax-M3 better for coding?

MiniMax-M3 scores higher on coding benchmarks: 41.5 versus 37.1 in the Noometry coding category.

### Which has the bigger context window?

MiniMax-M3 does, with 1M tokens against 131K.

### How many benchmarks do gpt-oss-20b and MiniMax-M3 share?

25 benchmarks have published results for both models. gpt-oss-20b has 34 scored results on Noometry and MiniMax-M3 has 41.