Mira Murati left OpenAI to start Thinking Machines. This week, the company released Inkling, a one-trillion-parameter open-weights multimodal model trained on forty-five trillion tokens, and uploaded it to Hugging Face under an Apache 2.0 license.

Open means anyone can download it, fine-tune it, run it on their own hardware. Multimodal means it speaks text, understands images, processes audio. The weights are public.

The model is not trying to be the strongest overall. Murati and team are clear on that. Inkling is built to be customizable. It uses a mixture-of-experts architecture, so only about forty-one billion parameters activate per task, making it efficient. Users can dial thinking effort up or down. It includes calibrated uncertainty, so it tells you when it is not sure instead of guessing.

What makes this different from other open models is who built it. You are looking at an OpenAI CTO who decided the frontier model race should have an open-weights option. The signal is that the smartest people at the frontier are ready to move away from the closed model arms race.

Inkling shipped with day-zero support in transformers, SGLang, and llama.cpp. People can use it today. This is not a research model. This is a product released by someone who knows how to build AI at scale.

——

Follow: @Ali Demi
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277

——

Sources:
https://techcrunch.com/2026/07/15/thinking-machines-amps-up-its-bet-against-one-size-fits-all-ai-with-its-first-open-model-inkling/
https://huggingface.co/blog/thinkingmachines-inkling
https://www.unite.ai/thinking-machines-lab-unveils-inkling-its-first-open-weights-multimodal-ai-model/

Repost this. Thanks.