Thinking Machines released Inkling, its first major language model under Apache 2.0 license. The 975 billion parameter MoE model is natively multimodal for text, images, and audio. It focuses on low cost with controllable thinking effort. The model emphasizes resistance to censorship while maintaining safety. Weights are available on Hugging Face. It performs strongly on software engineering and voice benchmarks. A smaller Inkling-Small version is also previewed. Founded by former OpenAI CTO Mira Murati, the company aims for customizable and open models. Inkling supports up to 1 million token context. This provides enterprises an option for on-premises or private cloud deployment with full control.
Thinking Machines released Inkling, its first major language model under Apache 2.0 license. The 975 billion parameter MoE model is natively multimodal for text, images, and audio. It focuses on low cost with controllable thinking effort. The model emphasizes resistance to censorship while maintaining safety. Weights are available on Hugging Face. It performs strongly on software engineering and voice benchmarks. A smaller Inkling-Small version is also previewed. Founded by former OpenAI CTO Mira Murati, the company aims for customizable and open models. Inkling supports up to 1 million token context. This provides enterprises an option for on-premises or private cloud deployment with full control.