unslothai/unsloth v0.1.902-beta: Command Palette + Desktop UI/UX
This release brings faster navigation, shareable run settings, and clearer errors to Unsloth Desktop.
ProofCode ↗
Key points
- Also, you can now share your GGUF model's run settings
- NVFP4 LoRA with 45% lower peak memory on Qwen3.8-27B-NVFP4
- INT4 + MXFP4 checkpoints stay packed during LoRA, including Kimi-K2.7-Code and gpt-oss
- Dense compressed-tensors INT4 checkpoints train as published on NVIDIA GPUs.
Sources (1)
- [1]unslothai/unsloth v0.1.902-beta: Command Palette + Desktop UI/UXGitHub: unslothai/unsloth · Oct 1, 02:06 PM
This release brings faster navigation, shareable run settings, and clearer errors to Unsloth Desktop.
Also, you can now share your GGUF model's run settings
Extractive summary: sentences quoted from the sources.
Hype check
Independently evaluated3 independent looks since
- Evaluation · Oct 3, 2026Qwen3.8 Flash Next, Swift 1.5, and MiMo V2.6: What I’m Testing Now
- Evaluation · Oct 6, 2026Qwen3.8 Flash Next Reasoning Modes: Off vs Low vs Medium vs Xhigh
- Evaluation · Oct 6, 2026Does Steering Break Your Model? A Multi-Dimensional Evaluation Suite for LLM Steering Methods
- Follow-up · Oct 1, 2026Si Sheppard – How did a few hundred Spanish soldiers topple two empires?
- Follow-up · Oct 1, 2026Introducing Clef: our open-source decision models, and new RL fine-tuning platform
- Follow-up · Oct 1, 2026Show HN: Breadcrumb, record everything on your mac + context manager for AI
Before this
- Sep 29, 2026NVIDIA/TensorRT-LLM v1.3.0rc29
- Sep 28, 2026unslothai/unsloth v0.1.900-beta: Laya Decision Models + Library
- Sep 22, 2026vllm-project/vllm v0.30.0
- Aug 22, 2026sgl-project/sglang v0.5.18
- Aug 8, 2026sgl-project/sglang v0.5.17
- Jun 26, 2026sgl-project/sglang v0.5.14