unsloth Releases

Identical to v0.1.52-beta, focusing on the introduction of DeepSeek V4 Flash and Kimi K3 support via Dynamic GGUFs and optimized downloading.

Aug 4, 2026

Identical to v0.1.52-beta, focusing on the introduction of DeepSeek V4 Flash and Kimi K3 support via Dynamic GGUFs and optimized downloading.

Aug 4, 2026

Introduces support for DeepSeek V4 Flash and Kimi K3 models via Dynamic GGUFs with optimized downloading, while explicitly confirming the project is unaffected by the Keyv npm supply chain incident.

Aug 4, 2026

This release addresses security and validation issues in Flowise, specifically fixing session ID validation, MCP command allowlists, and cross-workspace authorization. It also includes various utility and workflow fixes.

Jul 29, 2026

This release introduces support for the Kimi K3 and DeepSeek V4 Flash models via Dynamic GGUFs, along with new Deep Research and Parallel Chat modes. It significantly improves AMD and Intel GPU support, adds DoRA training, and includes various fixes for MLX, export, and inference.

Jul 29, 2026

Following v0.1.51, this update adds support for DeepSeek V4 Flash alongside the existing Kimi K3 capabilities. It continues to enhance parallel chat, deep research, and hardware compatibility while maintaining the stability improvements from the previous version.

Jul 29, 2026

This release introduces support for the Kimi K3 MoE model, enables parallel chat sessions, and adds a Deep Research mode for local web-based analysis. It also significantly improves AMD and Intel GPU support and introduces DoRA training alongside various fixes.

Jul 29, 2026

This release focuses on dashboard enhancements and database stability for SafeLine-CE, adding dark mode and upgrading PostgreSQL. It also resolves several licensing and configuration bugs.

Jul 28, 2026

This update optimizes FalkorDB integration and improves the graphiti-core library with various dependency updates and bug fixes. It includes enhancements to fulltext search and CI/CD workflows.

Jul 27, 2026

Unsloth now supports local LLM training and inference on AMD GPUs across Windows, WSL, and Linux. The update brings voice dictation, improved web search capabilities, and safer agent controls with tool-call permissions.

Jul 20, 2026

This release introduces comprehensive support for AMD GPUs, enabling local LLM training and inference on Radeon and Instinct cards across Windows, Linux, and WSL. Users can achieve up to 2x faster training speeds with 70% less VRAM usage, alongside new tools for managing complex hardware setups and AI agents.

Jul 20, 2026

This update focuses on personalization, adding three color palettes, custom fonts, and seven new languages. It also introduces Inkling model support, Intel GPU Vulkan inference, and Dynamic NVFP4 models.

Jul 15, 2026

A major version update for SubdomainX introducing comprehensive new scanning features including a REST API server, TUI dashboard, and advanced fingerprinting capabilities.

Jul 10, 2026

A release for SafeLine-CE focusing on security management improvements, including a read-only console role and fixes for authentication and rule persistence.

Jul 10, 2026

An update for SubdomainX focusing on deployment automation, documentation migration to Nextra v3, and the removal of deprecated ThreatCrowd integration.

Jul 9, 2026

DeepSeek-V4-Flash is now supported with thinking toggles, and the release adds NVFP4 export capabilities. It features a smarter OpenAI-compatible API, improved RAG and file chat, and support for Japanese and Portuguese.

Jul 7, 2026

This release focuses on DeepSeek-V4 support, enhanced NVFP4/FP8 exporting capabilities, and a smarter Llama-swap API system with improved RAG and file chat features.

Jul 7, 2026

A Textualize release addressing stability issues by fixing crashes related to screen padding and parsing extended keys, while also updating keyboard shortcuts.

Jun 30, 2026

Flowise update featuring MCP server integration, UI redesigns, support for newer AI models like Claude Opus 4.7 and GPT-5.5, and new audio generation capabilities.

Jun 25, 2026

Unsloth adds support for GLM 5.2 and achieves 3x longer context lengths with a new auto-fit algorithm. The update introduces a new Hub page, Canvas chat artifacts, thread forking, and Cloudflare HTTPS tunnels.

Jun 18, 2026