MiniMax-M3.1-Flash goes open weights with day-0 NVIDIA NeMo fine-tuning up to 128 GPUs
MiniMax_AI · x · 2026-10-23
- MiniMax confirmed that "Space Bunny", which quietly served 200T+ tokens, was actually MiniMax-M3.1-Flash, and the weights are now open.
- The model is an update of MiniMax-M3: a 60-layer MoE vision-language model with MiniMax Sparse Attention (MSA), optimized for agentic coding.
- NVIDIA NeMo offers day-0 fine-tuning via NeMo AutoModel: it loads the mixed-precision checkpoint directly (NVFP4 routed experts, MXFP8 projections) and supports FSDP2 with expert, pipeline, and context parallelism, scaling to 128 GPUs.
Related event: MiniMax Open-Sources M3.1-Flash, a 428B-Parameter MoE Model(4 posts)→
More from Infra
- Agents that don't close connections on retry waste ~148 tokens per task, vLLM tests show — No-Peanut-6988 · 2026-10-23
- Ex-Googler with 35 Years in Software Pursues Local AI Coding and Music Video Generation — julianharris · 2026-10-23
- Intelligence too cheap to meter: Opus price down 4x while AA score jumps 21 to 58 — haider1 · 2026-10-23
- SK Hyper in talks to fund envisioned $350B AI data center buildout in South Korea — dinabass · 2026-10-23
- MiniMax M3.1 Flash hits >1K tok/s decode with day-0 SGLang support across NVIDIA and AMD — MiniMax_AI · 2026-10-23
- RadixArk caught paying KOLs to coordinate quote-posts of SGLang announcement — EugeneVinitsky · 2026-10-23