Thinking Machines releases Inkling 2 Small, which fine-tuned another model on its own in 4.5 hours for $426

Original title: Thinking Machines' new model fine-tuned another model by itself - 4.5 hours, $426, no human in the loop

Thinking Machines Lab announced Inkling 2 Small on November 5 with a demo where the model ran a full ML project autonomously: exploring data, writing its own evaluator, building training sets, running SFT and RL experiments on Tinker, and debugging along the way. In 4.5 hours and $426 of spend, the company reports the trained model more than doubled the base model's score on a time-series forecasting task. The model is a 594B mixture-of-experts with 17B active parameters, trained with RL on hundreds of thousands of long-horizon coding tasks; it is available now via Tinker Playground with closed weights, and the company says it intends to release full weights under Apache 2.0 next week after external safety testing and red-teaming, though that part is announced rather than released. The au…

The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.