Thinking Machines open sources multimodal AI model, Inkling
AI startup Thinking Machines has released Inkling, an open-source multimodal language model. The model handles text, image, and audio and is designed for low cost and resistance to censorship.
AI startup Thinking Machines, founded by former OpenAI CTO Mira Murati, has launched Inkling, its first major language model released under an open-source Apache 2.0 license. This allows enterprises to customize and run the model on-premises or in virtual private clouds.
Inkling is a natively multimodal Mixture-of-Experts (MoE) model with 975 billion total parameters, capable of reasoning across text, images, and audio. The company highlights a "controllable thinking effort" mechanism designed to balance performance and cost. A smaller version, Inkling-Small with 276 billion parameters, has also been announced for low-latency, cost-sensitive applications.
The model shows strong performance on benchmarks for software engineering and voice understanding, according to the company. A noted differentiator is its design to "answer directly on topics that may be subject to censorship," aiming to provide a more trustworthy option for enterprises concerned about factual outputs.
Inkling is available on Hugging Face and via Thinking Machines' Tinker API. It enters a competitive landscape of open-weight models, offering businesses an alternative for agentic AI workloads that require customization and control.