unslothai/unsloth v0.1.908-beta: NVIDIA Nemotron 3.5 + Meta Muse Spark Open
Run NVIDIA's new Nemotron 3.5 models and Meta's new Muse Spark Open locally in Unsloth.
ProofCode ↗
Key points
- New: text-diffusion LLMs, DeepSpeed training and better prompt caching for coding agents.
- NVIDIA Nemotron 3.5 Bolt: Run on 24 GB of memory.
- Run NVIDIA Nemotron 3.5 Super VL, a 120B model with 12B active parameters and up to 1M tokens of context, built for document understanding, computer use and long-running agents.
- By default, Unsloth runs Nemotron 3.5 Bolt with a DSpark drafter (about 1 to 2 GB more memory) and Super VL with its built-in MTP for faster output.
Sources (1)
- [1]unslothai/unsloth v0.1.908-beta: NVIDIA Nemotron 3.5 + Meta Muse Spark OpenGitHub: unslothai/unsloth · Oct 15, 06:50 PM
Run NVIDIA's new Nemotron 3.5 models and Meta's new Muse Spark Open locally in Unsloth.
New: text-diffusion LLMs, DeepSpeed training and better prompt caching for coding agents.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 15, 2026huggingface/trl v1.16.0
- Oct 15, 2026NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal Reasoning
- Oct 15, 2026StepFun Releases Step 5 Preview Open Weights: 600B MoE, 27B Active, 1M Context
- Oct 15, 2026Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio
- Oct 15, 2026Welcome FLUX.3 Image Klein in Diffusers
- Oct 14, 2026NVIDIA/TensorRT-LLM v1.4.0rc0
