Thinking Machines debuts open-weight Inkling AI model
Thinking Machines is making Inkling downloadable, fine-tunable, and deployable as an open-weights foundation model.
The model is a 975B-parameter Mixture-of-Experts system with 41B active at inference and a context window up to one million tokens. It was trained across text, images, audio, and video, and supports native text, image, and audio inputs. Developers can fine-tune it through Tinker or use full checkpoints for NVIDIA Blackwell systems. Thinking Machines is also previewing a smaller 276B-parameter Inkling-Small model aimed at lower-cost, lower-latency use. TestingCatalog's note
The model is a 975B-parameter Mixture-of-Experts system with 41B active at inference and a context window up to one million tokens. It was trained across text, images, audio, and video, and supports native text, image, and audio inputs. Developers can fine-tune it through Tinker or use full checkpoints for NVIDIA Blackwell systems. Thinking Machines is also previewing a smaller 276B-parameter Inkling-Small model aimed at lower-cost, lower-latency use. TestingCatalog's note
score 7