MiniMax ships M3.1 Flash: native multimodal MoE with 1M context and sparse attention throughout
KyeGomezB · x · 2026-10-23
MiniMax released its M3.1 Flash model on Hugging Face: a natively multimodal MoE targeting coding, agentic workflows, and computer use, with 1M-token context and sparse attention throughout. The released architecture diagram shows Block Max Pooling feeding a Top-K index that selects KV blocks, with per-query-group sparse attention masks.
Related event: MiniMax Open-Sources M3.1-Flash: 428B Multimodal MoE with 1M Context(8 posts)→
More from Models
- DeepSeek 4.1 Flash Hits ~80 tok/s Locally on DGX Station via LM Studio — MatthewBerman · 2026-10-23
- MiniMax M3.1 Flash launches day-zero on Modular Cloud, tops open models in cowork automation — MiniMax_AI · 2026-10-23
- Neolab ecosystem ships flurry of updates: $1B Series A, new coding model Flint — FinanceYF5 · 2026-10-23
- Recursive Flint benchmarks: matches Qwen3.8-27B on agentic tasks, weak on long context — ChengleiSi · 2026-10-23
- Redditor open-sources behavioral architecture of his Monday custom GPT with 30% VRAM-saving trick — Top-Koala5617 · 2026-10-23
- Goshen Labs teases Ras Class models built on Qwen3.8-27B, coding-focused variants — Rasmic · 2026-10-23