v4.5.2
The popular open-source UI for local LLMs gets a new name, critical bug fixes, and optimized portable packages.
The open-source project formerly known as 'text-generation-webui' has taken a major step with its v4.5.2 release, officially rebranding to 'TextGen'. This update from developer oobabooga focuses on stability and user experience, addressing over a dozen bugs that plagued previous versions. Key fixes include resolving Gemma-4's tool-calling issues with argument parsing, preventing chat scroll from getting stuck, and correcting token handling for models without chat templates. These improvements make running advanced local models like Gemma 2 or Llama 3.1 significantly more reliable for developers and researchers.
Beyond bug fixes, v4.5.2 delivers practical deployment enhancements. It updates core backends, pulling in the latest commits from llama.cpp and its fork ik_llama.cpp for improved performance and new quantization types. Most notably, the release provides a full suite of updated portable builds. These are self-contained packages (200MB-1.2GB) for GGUF models that require no installation—users simply download for their OS and GPU platform (NVIDIA CUDA, AMD ROCm, Vulkan, or CPU) and run. This dramatically lowers the barrier to entry for experimenting with local LLMs, turning what was often a multi-step setup into a one-click experience.
- Official project rename from 'text-generation-webui' to 'TextGen', with a new GitHub URL at github.com/oobabooga/textgen.
- Critical bug fixes for Gemma-4 tool calling, chat UI rendering, and BOS/EOS token handling across various backends like ExLlamaV3.
- New portable, no-install builds for Windows, Linux, and macOS supporting multiple GPU backends (CUDA, ROCm, Vulkan) and updated llama.cpp dependencies.
Why It Matters
This release stabilizes a key tool for the local LLM ecosystem, making advanced model experimentation more accessible and reliable for developers and AI enthusiasts.