Thinking Machines releases Inkling 2 Small on Tinker: a 594B-parameter multimodal model, with open weights planned

On November 5, 2026, Thinking Machines Lab released Inkling 2 Small, the first model of its second Inkling generation. It is a Mixture-of-Experts model with 594B total parameters and 17B active per token. It accepts text and image input, writes text, supports variable reasoning effort and, according to its model card, handles a context window of up to 1M tokens (on Tinker, the largest version offered has a 256K context). It does not support audio.

At launch the model is available only through Thinking Machines' Tinker platform, for sampling and fine-tuning, and in Tinker Playground, with closed weights under the Tinker Terms of Use. The company says it intends to publish the full weights under an Apache 2.0 license next week, after external safety testing and red-teaming, and to bring the model to other platforms in the coming weeks. Tinker lists $0.30 per million input tokens ($0.03 cached) and $1.00 per million output tokens; training costs $1.50 per million tokens on the 64K-context version and $3.00 on the 256K version.

Artificial Analysis scored it 31 on its Intelligence Index v5.0.1, 18 points above Inkling Small, while noting that it uses about four times as many output tokens per task and costs $1.18 per index task. As of November 5, 2026, Artificial Analysis said the planned weights release would make it the highest-scoring open-weights model developed outside China. Thinking Machines says it trained the model with reinforcement learning on long coding and research tasks, and lists limits: it is not a concise communicator, is less tuned for chat and can over-react to safety-related questions.