Two new models, MiniMax-M3.1-Flash and incoai/Qwen3.8-Flash-Next-DFlash2, have been announced. The MiniMax model is from MiniMax AI, while the Qwen model is from the original authors of DFlash 2 and claims to be 20% faster than MTP at concurrency 1. AI
IMPACT New open-source models offer potential speed improvements for local LLM deployments.
RANK_REASON Two new open-source models are announced, with one claiming a speed improvement.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →