Thinking Machines says its new Inkling announcement brings a roughly 1 trillion-parameter, 41B-active Mixture-of-Experts model into the open, with native text, image and audio input and full weights available through its Hugging Face model card. In a widely shared post on X, researcher Nathan Lambert called the benchmarks "a clear step up from Nemotron Ultra," while also saying Inkling appears a bit behind GLM 5.2 on some agentic tests and Kimi K 2.6 on multimodal work.