Meet Inkling. Think out loud with it.
Inkling is Thinking Machines Lab's new open-weights model — a 975B-parameter Mixture-of-Experts generalist with a 1M-token context, native audio and vision, and reasoning you can dial up or down. We've got it running live on Gentic infrastructure, so you can talk to it right here.
Try it live
Send a message and watch Inkling reason before it answers. Pick how hard it thinks.
Ask Inkling anything. It streams its reasoning first, then the answer.
Running Inkling-NVFP4 live on Gentic infrastructure · responses are capped and rate-limited
Why Inkling is worth a look
Thinking Machines built Inkling as a customizable foundation — not a race for the single strongest model, but a well-rounded base you can make your own.
Controllable reasoning
Dial thinking effort up or down to trade speed for depth. Inkling matches Nemotron 3 Ultra on Terminal Bench 2.1 at roughly a third of the tokens.
Natively multimodal
Pretrained on 45 trillion tokens of text, images, audio, and video — at the open-source frontier for speech.
Open weights, ready to tinker
A well-rounded generalist built as a base for customization — open weights on Hugging Face and fine-tunable on Tinker.
Inkling and the Inkling name are the work of Thinking Machines Lab. This playground runs the open-weights model on Gentic infrastructure for demonstration.