Meta has released Muse Glimmer, an open source, local, agentic, and multimodal 30B parameter model featuring a 2B ViT-style vision encoder and a 28B parameter text decoder. Day-0 support is available across multiple libraries, including transformers, llama.cpp, vLLM, and Inference Endpoints.
Baseten has become a supported inference provider on the Hugging Face Hub, allowing developers to run open-weight language models directly from model pages, client SDKs, and agent harnesses. The initial integration brings support for conversational and text-generation tasks using models such as Kimi K3, DeepSeek V4 Flash, and GLM-5.2.
Liquid AI has released LFM2.5-2.6B, a new model built for local agent deployment on edge devices. Pre-trained on approximately 34T tokens with a 128K context window, the model achieves decode speeds of 220 tokens per second on an Apple M5 Max and features day-one ecosystem support.