Thinking Machinesintroduced Tinker on October 1, 2025, to let researchers control a model's training data and learning procedure while outsourcing the distributed computing infrastructure. Its API exposes training and sampling operations, while the service handles scheduling, resource allocation and recovery from failures.
That division lets researchers write their own training logic without operating the underlying cluster. The launch described a private beta; the later general-availability release removed the waitlist on December 12, 2025.
Terms in this piece · Glossary
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
LoRA — A cheap way to fine-tune a model by training a small add-on layer instead of changing all of the model's weights.
