Inkling is a general-purpose multimodal AI model developed by Thinking Machines Lab. As an open-weight large language model, it is designed to provide flexible deployment, customization, and development options for researchers and developers.
Built with a Mixture-of-Experts (MoE) architecture, Inkling features approximately 975B total parameters with around 41B active parameters and supports ultra-long context processing of up to 1 million tokens.
Inkling natively supports text, image, and audio inputs with text generation capabilities. It is optimized for multimodal reasoning, AI Agent workflows, tool calling, software development, and intelligent application creation.
With open model weights, developers can deploy, fine-tune, and customize Inkling for specific use cases, including coding assistants, AI chatbots, RAG knowledge systems, enterprise automation, and multimodal AI solutions.
Inkling Small
15 Thinking Machines has released a smaller version of Inkling called Inkling-Small.
“Small” is relative here: it still has 276B total parameters and 12B active parameters. The interesting part is that, despite being roughly a quarter the size of Inkling, it actually scores higher on reasoning and agentic coding.
Thinkingmachines
Comments (0)