# GPT-4 vs MiniMax-M3

> MiniMax-M3 is the stronger model overall, scoring 43.8 to 29.1 on the Noometry Index.

- Canonical page: https://noometry.com/compare/gpt-4-vs-minimax-m3
- Last updated: 2026-10-10
- Shared benchmarks: 25

## Summary

- They share 25 benchmarks with published results for both. GPT-4 scores higher in 0 categories and MiniMax-M3 in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where MiniMax-M3 leads 58.4 to 18.4.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.1% for GPT-4 and 71.1% for MiniMax-M3.
- MiniMax-M3 is cheaper at $0.30 / $1.20 per million input/output tokens, against $30 / $60 for GPT-4.
- MiniMax-M3 accepts more context: 1M tokens versus 8K.
- MiniMax-M3 has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-4 | MiniMax-M3 |
|---|---|---|
| Provider | OpenAI | MiniMax |
| Noometry Index | 29.1 | 43.8 |
| Rank | 316 | 85 |
| Context | 8K | 1M |
| Input $/M | $30 | $0.30 |
| Output $/M | $60 | $1.20 |
| Weights | Proprietary | Open |

## Coding

- GPT-4: 31.6 (#283)
- MiniMax-M3: 41.8 (#118)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| LMArena Coding | 1254 | 1469 |
| FrontierCode | — | 14.7% |
| LMArena WebDev | — | 1482 |
| SciCode | — | 47.1% |
| WeirdML | 12.4% | — |
| BigCodeBench Instruct | 46% | — |
| BigCodeBench Complete | 57.2% | — |
| ALE-Bench | — | 640.02 |
| HumanEval+ | 79.3% | — |

## Agentic & Tool Use

- GPT-4: —
- MiniMax-M3: 22.6 (#130)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| APEX-Agents | — | 37.7% |
| OSWorld 2.0 | — | 4.6% |
| GBAEval | — | 0.9% |
| METR Time Horizons | 36.1% | — |
| Vending-Bench 2 | — | 2,158 |

## Reasoning

- GPT-4: 17.8 (#289)
- MiniMax-M3: 30.1 (#87)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| Chess Puzzles | 4% | 14% |
| LMArena Hard Prompts | 1241 | 1447 |
| Mystery Game Puzzles | 12% | 8% |
| DTBench | 62.7% | 78.9% |
| LMCA | 17.1% | 33.7% |
| Epoch Capabilities Index | 125.89 | 146.95 |
| ForecastBench | 57.8 | 61.4 |
| SimpleBench | — | 45.8% |
| NYT Connections (extended) | — | 65.1% |
| CritPt | — | 3.7% |
| Surface Evolver Bench | — | 55% |
| BIG-Bench Hard | 75.1% | — |
| HellaSwag | 95.3% | — |
| WinoGrande | 87.5% | — |

## Math

- GPT-4: 10.8 (#309)
- MiniMax-M3: 40.0 (#95)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 71.1% |
| LMArena Math | 1269 | 1429 |
| ProofBench | — | 18% |
| MATH Level 5 | 23% | — |
| GSM8K | 92% | — |

## Knowledge

- GPT-4: 18.4 (#282)
- MiniMax-M3: 58.4 (#35)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| GPQA Diamond | 35.7% | 90.9% |
| LMArena Expert | 1211 | 1461 |
| MMLU | 86.4% | — |
| TriviaQA | 84.8% | — |

## Multimodal

- GPT-4: —
- MiniMax-M3: 40.2 (#51)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| LMArena Vision | — | 1253 |
| LMArena Document | — | 1435 |

## Multilingual

- GPT-4: 40.6 (#215)
- MiniMax-M3: 53.0 (#75)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| LMArena Non-English | 1246 | 1420 |
| LMArena Chinese | 1242 | 1463 |
| LMArena French | 1283 | 1447 |
| LMArena German | 1251 | 1426 |
| LMArena Japanese | 1209 | 1381 |
| LMArena Korean | 1184 | 1372 |
| LMArena Russian | 1251 | 1428 |
| LMArena Spanish | 1261 | 1432 |

## Instruction Following

- GPT-4: 65.3 (#222)
- MiniMax-M3: 75.5 (#62)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| LMArena Instruction Following | 1241 | 1433 |

## Long Context

- GPT-4: 37.7 (#212)
- MiniMax-M3: 44.2 (#72)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| LMArena Longer Query | 1244 | 1445 |

## Writing & Preference

- GPT-4: 34.9 (#268)
- MiniMax-M3: 62.1 (#83)

| Benchmark | GPT-4 | MiniMax-M3 |
|---|---|---|
| LMArena Text | 1263 | 1433 |
| LMArena Creative Writing | 1244 | 1404 |
| LMArena Multi-Turn | 1257 | 1442 |
| EQ-Bench Creative Writing | 752 | — |
| EQ-Bench 4 | — | 1150 |

## FAQ

### Is GPT-4 better than MiniMax-M3?

MiniMax-M3 is the stronger model overall, scoring 43.8 to 29.1 on the Noometry Index.

### Which is cheaper, GPT-4 or MiniMax-M3?

MiniMax-M3 is cheaper. It lists at $0.30 per million input tokens and $1.20 per million output tokens; GPT-4 lists at $30 and $60.

### Is GPT-4 or MiniMax-M3 better for coding?

MiniMax-M3 scores higher on coding benchmarks: 41.8 versus 31.6 in the Noometry coding category.

### Which has the bigger context window?

MiniMax-M3 does, with 1M tokens against 8K.

### How many benchmarks do GPT-4 and MiniMax-M3 share?

25 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and MiniMax-M3 has 41.