Thinking Machines Lab has launched Inkling, a groundbreaking AI model featuring 975 billion parameters. The model utilizes a Mixture-of-Experts architecture, activating only 41 billion parameters at a time for efficiency.
Details on Inkling's Architecture
Inkling is designed to handle multiple data types, including text, audio, and video, while supporting a context window of up to 1 million tokens. The use of Mixture-of-Experts technology allows the model to manage compute costs while maintaining high performance levels. Alongside Inkling, a smaller variant named Inkling-Small has been introduced, containing 12 billion active parameters.
Comparison with Competitors
Founded by former OpenAI CTO Mira Murati in February 2025, Thinking Machines Lab has positioned itself differently from companies like OpenAI and Anthropic, which restrict access to their most advanced models through APIs. The startup, having raised $2 billion, aims to provide customizable AI solutions for enterprises. This approach allows businesses to run models on their own infrastructure without sharing sensitive data with third parties.
Inkling’s launch does not yet seem to have affected market reactions significantly, as Thinking Machines is still a private entity. However, its placement in the market challenges established players like Meta, known for its Llama series, and Mistral, which has a foothold in Europe. The model's release under the Apache 2.0 license enables developers to download and modify it freely, raising concerns about monetization strategies moving forward.
This material is informational and not financial advice.
