AI BriefingInkling-2-Small: 383B MoE multimodal model deployed at 8-bit precision
thinkingmachines/Inkling-2-Small
About the project
Inkling-2-Small, released by Thinking Machines Lab, is a multimodal LLM that understands both images and text simultaneously. It is based on a Mixture of Experts (MoE) architecture with a total of 383B parameters and is stored at 8-bit precision, reflecting a configuration designed for memory efficiency.
Given an image URL and a text prompt as input, it generates responses based on visual information. Beyond simple image description, it supports tool calling and structured output, making it suitable for agent-based workflows. It does not support audio input.
Inference is available immediately through the Hugging Face Transformers library and vLLM servers. It is also provided live by major inference providers such as DeepInfra and Together AI, enabling immediate access via cloud APIs in addition to local environments.
Released under the Apache 2.0 license, it can be freely used in commercial projects. It is suitable for developers who require large-scale multimodal inference but prefer a lighter deployment environment over full-precision models.
thinkingmachines/Inkling-2-Small
The original page has no description.
image-text-to-text
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.