Inkling 2 Small Arrives on Tinker, With Open Weights Still to Come
Thinking Machines Lab’s new multimodal model is available through Tinker. Its planned Apache 2.0 weight release still depends on external safety testing.
The actual story.
Thinking Machines Lab introduced Inkling 2 Small on November 5, starting access through Tinker rather than immediately releasing its weights. The company says an Apache 2.0 release is intended for the following week, after external safety testing and red-teaming. That is a plan, not a completed release.
What “Small” describes
The mixture-of-experts model has 594 billion total parameters, with 17 billion active. It accepts image inputs and offers adjustable effort. The name should not be read as a promise that the full model fits on an ordinary laptop.
What developers can test now
The company reports an Artificial Analysis Intelligence Index score of 31 and improvements in coding, vision and reasoning. These reported results do not guarantee performance on a particular workflow. Tinker provides sampling and fine-tuning; downloadable checkpoints are tied to the weights release.
The model card describes text and image inputs with text output. Teams evaluating it should distinguish hosted access today from the separate question of running the released model themselves.
Test the available model against your own tasks; treat the announced weights release as pending.
Receipts, not rumors.
Read the original reporting: Thinking Machines Lab
The reporting above provides context. Illustrations are editorial interpretations, not source photographs.
What’s your take?