Former OpenAI CTO Releases ‘Inkling’ : A Truly Open Frontier AI Model

3 1 minute read

In a move that has sent shockwaves through the tech industry, Mira Murati, former CTO of OpenAI, has officially unveiled her latest venture’s debut: Inkling. Released by her new startup, Thinking Machines Lab, Inkling marks a historic moment as the largest American open-weight AI model released to date.

A New Standard for Open Weights

Inkling is a massive 975-billion parameter Mixture-of-Experts (MoE) model. Released under the Apache 2.0 license, it offers a level of accessibility rarely seen in frontier-class models. Unlike closed-source rivals, Inkling allows developers to download and customize the weights for their specific needs.

A close-up of high-end server hardware, representing the massive computing power required to run Inkling.

The model features 256 routed experts, with six experts active per token generation. This architecture ensures high performance while maintaining efficient inference. However, with its sheer size, running the full version requires significant hardware, typically exceeding 2 TB of GPU memory.

Performance and Accessibility

Thinking Machines Lab is positioning Inkling to compete directly with global leaders like DeepSeek V4 and GLM 5.2. For those with more modest hardware, the lab also announced Inkling-Small, a 276B MoE model with 12B active parameters designed for lower latency and more accessible deployment.

A minimalist graphic banner with clean typography celebrating the release of the Inkling frontier model.

"Our goal is to provide a truly open alternative for organizations that want to own their intelligence," the lab stated during the launch. Both models are currently available on the Tinker platform and feature full integration with Hugging Face.

A developer working on a tablet, showcasing the practical application and accessibility of smaller AI models like Inkling-Small.

Why This Matters

For the community-focused readers at Brownstone Worldwide, this release signals a shift toward more democratic AI. By providing open weights, Thinking Machines Lab is empowering a broader range of creators and businesses to innovate without the restrictive "walled gardens" of traditional AI giants.

A digital shield representing AI data security and the protection offered by open-weight enterprise models.

As digital media and news publishing continue to evolve, tools like Inkling will likely play a pivotal role in how we generate and verify content in the near future.

A professional working at a desk, illustrating the integration of advanced AI tools into daily business operations.