v0.1.501-beta

✨ New Features

• Local LLM training and inference on AMD GPUs across Windows, WSL, and Linux.
• Support for RDNA2 and Gorgon Halo GPUs.
• Optimized ROCm builds supporting GGUF & Safetensors inference.
• Preliminary support for voice dictation / whisper.cpp for fast text to speech.
• Unified memory safetensors loading is 2x faster.
• Gradient checkpointing is much faster for unified memory devices.
• Automatic GPU placement and manual GPU/layer selection for running larger models.
• Ability to move MoE expert layers into system memory.
• Splitting models across multiple GPUs or using Tensor Parallelism.
• Saving hardware settings separately for each model and quant.
• Resume long chats without rebuilding the full conversation after an idle model frees its VRAM.
• Stalled Hugging Face XET downloads automatically retry over standard HTTP.
• Existing GGUF files are reused instead of being downloaded again when a model loads.
• Web search can now read PDF papers, manuals, and other PDF results.
• Parallel tool calls, reasoning output, and tool retries work more reliably.
• Opt-in MCP endpoint for compatible AI clients to inspect models, training history, start/stop training, load checkpoints, validate recipes, and export GGUFs.
• Dynamic NVFP4 models (keeps accuracy-sensitive layers in FP8 or BF16 while running the rest in W4A4) for faster inference and longer context on NVIDIA Blackwell GPUs.
• Personalization options for Studio: Standard, Classic, and Minimal color palettes (light/dark modes), custom colors, font import, font size adjustment, contrast, reduced motion.
• Voice settings tab in Studio for dictation, custom dictionary settings, and read-aloud.
• Unsloth Studio available in French, German, Spanish, Hindi, Arabic, Russian, and Korean.
• Four-level tool-call permission selector for agents: Ask, Approve for me, Off, and Full access.
• Agent workspace isolation and safer installer checks.