Models & Hardware
MiniMax’s Own Table Puts DeepSeek-1017 Ahead on Coding
77.0 to 73.7 on DeepSWE; MiniMax wins Automation-Bench. @MiaAI_lab ranks DeepSeek first of what you can run.
The vendor put a rival that beats it in its own table.
@MartinSzerment read MiniMax’s benchmark table closely. On coding, DeepSeek’s newest Flash wins. On BankerToolBench, M3.1 Flash scores 76.2, the same as Opus 5.
| Benchmark | MiniMax-M3.1-Flash | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| DeepSWE v1.1 | 73.7 | 77.0 |
| Terminal Bench 4.0 | 38.1 | 42.3 |
| NL2Repo-Bench | 61.2 | 68.2 |
| Automation-Bench v1.0.6 | 62.8 | 60.5 |
| HLE (without tools) | 43.4 | 42.6 |
@nuonrg says Artificial Analysis has M3.1 Flash at 42, on par with GLM-5.3-Flash.
For local owners, @MiaAI_lab gives her ranking in a reply: “Among the model we can realistically run, Qwen is only inferior to DeepSeek V4.1 Flash 1017.”
Sources
- @MartinSzerment on X · Oct 22, 9:59 PM PT
- @nuonrg on X · Oct 23, 12:51 AM PT
- @MiaAI_lab on X · Oct 23, 6:47 AM PT
- MiniMax-M3.1-Flash model card · huggingface.co