Thinking Machines releases Inkling 2 Small with Apache 2.0 license
Thinking Machines launched Inkling 2 Small, a 594 billion parameter Mixture of Experts model, which is expected to be released as open weights under the Apache 2.0 license within a week. The release follows a nine-day reinforcement learning phase on 2,000 GPUs and positions the new lab to compete through strategic open-sourcing.
Thinking Machines has launched Inkling 2 Small, a 594 billion parameter Mixture of Experts model available on Tinker. The model is expected to be released under the Apache 2.0 license as open weights within the next week, following a red teaming process. The main reinforcement learning phase for the model took nine days on 2,000 GPUs, covering 1,000 steps and more than 17 million rollouts.
Benchmark results for Inkling 2 Small include 86.3 percent on SWE bench pro v2 hard, 81 on MMMU pro, and 70 on Job Bench. Business Insider reported that Thinking Machines is looking to raise approximately $5 billion at a $40 billion pre-money valuation, though this funding round is not yet finalized. The release is viewed as impressive for a new lab, with open-sourcing considered a strategic necessity for companies not at the cutting edge, similar to the approach taken by Meta. As one observer noted, "I think that's the only way to grow."