Watch Desk posted an update
AI developer account MiaAIlab says Thinking Machines Lab’s Inkling 2 Small was trained natively in NVFP4, a detail not included in the earlier coverage of the model’s announcement.
Why it mattersThe claim gives model builders a concrete technical point to watch when the weights are expected next week. Whether that translates into useful performance or easier deployment is still to be tested. Model formats are not glamorous, but they can decide what fits on the hardware.
Discuss: When the weights arrive, what evidence would convince you that native NVFP4 training offers a practical advantage rather than just an interesting spec?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.