v0.1.702-beta

✨ New Features

• Added tool calling / web search & more for all external providers.
• UI and UX fixes - VRAM usage is now tunable.
• 10% faster inference + reduced VRAM usage and other perf fixes.
• Much better AMD RDNA3,4 + Strix Halo, Mac support.
• Image diffusion, video generation fixes.
• Can login with Codex subscription.
• Added self-healing tool calling that repairs malformed calls.
• Models can run Python and Bash inside sandboxed environments.
• Added unlimited private web search, Deep Research, RAG and MCP.
• Local models can be served through Unsloth's OpenAI-compatible API.
• Added support for running Muse Glimmer 30B, Kimi K3, Qwen3.8, DeepSeek-V4 Flash 0731, Gemma 4.
• Added video generation with MiniMax-H3.
• Added ability to create images and videos with diffusion models at up to 2x faster inference.
• Added no-code training for LLMs, diffusion models, TTS models and embedding models.
• Training is up to 2x faster and uses up to 70% less VRAM on supported LLM workloads.
• Added ability to export models to NVFP4, GGUF and other formats.
• Added ability to securely deploy and access models remotely via Cloudflare HTTPS.
• Added support for Windows, macOS, and Linux.
• Added support for CPU and multi-GPU systems, NVIDIA and AMD GPUs, Intel hardware, and Mac.
• Added CPU support for Chat and Data Recipes.