MiniMax has released its M3.1 Flash model, which has demonstrated strong performance in complex Agent tasks, outperforming DeepSeek's V4.1 Flash-1017 in this area. While DeepSeek maintains an edge in software engineering tasks, MiniMax's new model excels in scenarios requiring multi-step tool calls and interaction across systems, as evidenced by its top ranking on the Automation-Bench dataset. The M3.1 Flash also offers competitive pricing, making it an attractive option for developers seeking both capability and cost-effectiveness. AI
IMPACT This release intensifies competition in the Flash model segment, offering developers a more capable and cost-effective option for complex Agent tasks.
RANK_REASON New model release from a frontier lab (MiniMax) with performance benchmarks. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Artificial Analysis
- BankerToolBench
- Claude Opus 5.5
- DeepSeek
- DeepSWE v1.1
- GPT 5.6 "Sol"
- Hassan
- M3.1 Flash
- MiniMax
- OpenCode
- Zapier
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →