AI CURATED HOT STORY · 10/19/2026
MiniMax has officially released its new lightweight model, M3.1 Flash, which previously dominated leaderboards on platforms like OpenRouter and OpenCode under the anonymous codename "Space Bunny," processing over 56.7 trillion tokens. This launch marks a significant breakthrough in agentic tasks, code generation, and computer use, with performance metrics ranking first among domestic models in several key areas.
Cross-verified data from Artificial Analysis and Arena.ai confirms that M3.1 Flash achieves top-tier capabilities while optimizing efficiency. The model's intelligence index score increased by 13 points, and it ranked 12th among open-source models in Code Arena WebDev, with a total score jump of 106 points compared to its predecessor. Crucially, inference speed improved nearly threefold, while usage costs dropped by 50-70% due to optimized token output, offering highly competitive API pricing.
The release significantly impacts the developer ecosystem. M3.1 Flash is now available via API and coding platforms, with official weights scheduled for open-source release on October 23. This combination of high performance, low cost, and open-source availability provides a cost-effective solution for complex tasks in high-frequency scenarios, intensifying competition in the lightweight LLM market.