Thinking Machines Announces Inkling 2 Small With Weights to Follow Under Apache-2.0
The 594B/17B MoE model is live on Tinker; full weights are announced to follow under Apache-2.0 after external safety testing.
Thinking Machines announced Inkling 2 Small on 5 November 2026, a mixture-of-experts language model with 594 billion total parameters and 17 billion active parameters, supporting image input and a variable-effort inference mode. The company states the model was trained natively in NVFP4 precision on GB300 NVL72 systems.
The company reports a score of 31 on the Artificial Analysis Intelligence Index, up from 13 for the previous generation; this figure is a vendor claim, as the model weights are not yet publicly available for independent evaluation. The model is accessible through the company's Tinker platform for inference and fine-tuning, with a listed price of USD 0.30 per million input tokens and USD 1.00 per million output tokens and a default context window of 64K tokens.
Full model weights are announced to follow under the Apache-2.0 license after external safety testing and red-teaming. As of the announcement date, the weights had not been released.