Welcome Inkling by Thinking Machines
Inkling by Thinking Machines is a large, open-source multimodal LLM with 1T parameters and 1M context window, designed to handle image, text, and audio inputs, trained on 45 trillion tokens across multiple modalities. It uses a decoder-only MoE architecture, supports inference via SGLang, llama.cpp, and vLLM, and offers low token consumption and efficient multimodal reasoning. The model excels in tasks like document processing and complex reasoning, with features such as MTP and compatibility with agentic workloads.
More from Huggingface
LFM2.5-VL-3B for Better and Faster Vision Capabilities for the Edge
Thinking of ACE? We Can Do It with Fewer Tokens
Meta is back with Muse Glimmer: local, agentic, multimodal, and open source
