MiniMax-M2 Instruct vs Qwen3: which model should you choose?
We compared MiniMax-M2 Instruct and Qwen3 on price, context window and 12 benchmarks. Qwen3 has the stronger average benchmark score; Qwen3 is cheaper on a blended 3:1 basis.
MiniMax-M2 Instruct
MiniMax-M2 Instruct is a multimodal model from MiniMax. It offers a 128,000-token context window, costs $5.11 per million input tokens and $25.04 per million output tokens, and averages 79.4 across the index's benchmark suite.
- Input
- $5.11 / 1M
- Output
- $25.04 / 1M
- Context
- 128,000
- Avg score
- 79.4
- Open weights
- Yes
Qwen3
Qwen3 is a multimodal model from Alibaba Cloud. It offers a 128,000-token context window, costs $5.04 per million input tokens and $22.52 per million output tokens, and averages 80.3 across the index's benchmark suite.
- Input
- $5.04 / 1M
- Output
- $22.52 / 1M
- Context
- 128,000
- Avg score
- 80.3
- Open weights
- Yes
Specification comparison
| Spec | MiniMax-M2 Instruct | Qwen3 | Edge |
|---|---|---|---|
| Input price / 1M | $5.11 | $5.04 | B |
| Output price / 1M | $25.04 | $22.52 | B |
| Context window | 128,000 | 128,000 | Tie |
| Max output | 4,096 | 8,192 | B |
| Average score | 79.4 | 80.3 | B |
| Arena Elo | 1442.2 | 1420.9 | A |
| Open weights | Yes | Yes | Tie |
Benchmark comparison
MiniMax-M2 Instruct wins 5, Qwen3 wins 7, and 0 benchmarks are tied. The table shows the modelled score for each model and the gap.
| Benchmark | MiniMax-M2 Instruct | Qwen3 | Gap | Winner |
|---|---|---|---|---|
| MMLU-Pro | 82.1 | 81.2 | 0.9 | MiniMax-M2 Instruct |
| GPQA Diamond | 73.7 | 77.7 | 4.0 | Qwen3 |
| SWE-bench Verified | 69.9 | 71.6 | 1.7 | Qwen3 |
| HumanEval | 86.7 | 87.0 | 0.3 | Qwen3 |
| MATH-500 | 82.1 | 86.5 | 4.4 | Qwen3 |
| AIME 2025 | 77.8 | 74.5 | 3.3 | MiniMax-M2 Instruct |
| LiveCodeBench | 71.1 | 73.8 | 2.7 | Qwen3 |
| IFEval | 87.3 | 85.6 | 1.7 | MiniMax-M2 Instruct |
| TruthfulQA | 80.0 | 80.4 | 0.4 | Qwen3 |
| HellaSwag | 88.5 | 88.1 | 0.4 | MiniMax-M2 Instruct |
| MMMU | 74.1 | 76.7 | 2.6 | Qwen3 |
| Arena Elo | 1442.2 | 1420.9 | 21.3 | MiniMax-M2 Instruct |
Price analysis
On a blended 3:1 input-to-output basis, MiniMax-M2 Instruct costs about $10.09 per million tokens and Qwen3 costs about $9.41. Qwen3 is roughly 7% cheaper than MiniMax-M2 Instruct. At high volume, that gap compounds quickly, so price should carry real weight unless the quality difference is decisive.
Both models list a 128,000-token context window. That is enough for long documents, large codebases and multi-step agent traces.
Verdict
Across 12 shared benchmarks, MiniMax-M2 Instruct wins 5, Qwen3 wins 7, and 0 are tied. On the index average, Qwen3 leads MiniMax-M2 Instruct by 0.9 points (80.3 versus 79.4).
The price picture favours Qwen3. The context picture is effectively a tie. Those two dimensions often matter more than a few benchmark points once a model is good enough for the task.
If you need a single default, start with Qwen3 for quality-sensitive work, and check whether its price holds at your expected volume before committing.
Choose MiniMax-M2 Instruct if
- You prefer MiniMax's tooling or already use it.
Choose Qwen3 if
- You want the higher average benchmark score in this index.
- Unit cost matters and you want the cheaper blended price.
- You need the larger 8,192-token output limit.
FAQ
Is MiniMax-M2 Instruct better than Qwen3?
On the index average, Qwen3 leads MiniMax-M2 Instruct by 0.9 points (80.3 versus 79.4). The better choice depends on price, context and your task. MiniMax-M2 Instruct wins 5 benchmarks, Qwen3 wins 7.
Which is cheaper, MiniMax-M2 Instruct or Qwen3?
Qwen3 is cheaper on a blended 3:1 basis at about $9.41 per million tokens, versus $10.09 for MiniMax-M2 Instruct.
Which has the larger context window?
Both models list a 128,000-token context window.
Are these models open weights?
MiniMax-M2 Instruct is open weights; Qwen3 is open weights.
Which should I choose for production?
If you need a single default, start with Qwen3 for quality-sensitive work, and check whether its price holds at your expected volume before committing.