PulseAugur
EN
LIVE 20:02:15

Unsloth Studio releases v0.1.48-beta with enhanced model export and API features

Unsloth Studio has released version 0.1.48-beta, introducing significant enhancements to model exporting and API serving capabilities. The update allows for exporting models in various formats including NVFP4, FP8, and imatrix GGUFs, alongside improved support for GGUF LoRA and source-matched exports. Additionally, Unsloth now offers a more robust OpenAI-Compatible API serving with safer automatic model swapping, idle auto-unload for VRAM management, and enhanced tool-call healing. The release also brings improvements to RAG and file chat functionalities, better PDF and Word document handling, and a customizable embedding model. AI

IMPACT Enhances efficiency and flexibility for developers working with local AI models and APIs.

RANK_REASON This is a software release for a tool that enhances AI model operations, not a core AI model release from a frontier lab.

Read on Unsloth — Releases →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Unsloth Studio releases v0.1.48-beta with enhanced model export and API features

COVERAGE [1]

  1. Unsloth — Releases TIER_1 Nederlands(NL) · shimmyshimmer ·

    DeepSeek-V4 + NVFP4 Exporting

    <p>Unsloth Studio can now export to NVFP4, FP8, imatrix GGUFs after training, can act as a llama-swap API system, includes Japanese and Brazilian language support, includes MLX, safetensors tool calling + healing support and much more! Unsloth core makes GRPO 1.3x faster, have HT…