Ex-OpenAI CTO Mira Murati’s Thinking Machines Launches Its First Open-Weight AI Model
With 975 billion total parameters, Inkling is one of the largest models of its kind.
News
- U.S. Relaxes AI Chip Curbs for UAE Following Support in Iran Conflict
- Ex-OpenAI CTO Mira Murati’s Thinking Machines Launches Its First Open-Weight AI Model
- AI Vendor Lock-In Emerges as the Next Enterprise Risk in the UAE: IBM Study
- Meta Sued for Allegedly Using AI to Lay Off Employees on Leave
- Hundreds of AI Experts Urge Governments to Prepare for AI-Driven Economic Disruption
- Saudi Arabia Adds Agentic AI to Its National Data Platform
Mira Murati’s AI lab Thinking Machines has unveiled a new artificial intelligence model that could potentially be an alternative to popular open-source offerings from Chinese AI labs.
The model, named Inkling, is the lab’s first in-house AI model, which, unlike its competitors from OpenAI, Anthropic, and Google, is open-weight, meaning external developers and companies can download and modify it, unlike proprietary, closed-source models.
With 975 billion total parameters, Inkling is one of the largest foundational models of its kind. It is also trained on 45 trillion tokens of text, image, audio, and video. While acknowledging Inkling may not be the strongest overall model available today, it is trained to be a “broad, balanced foundation model.”
Along with Inkling, Thinking Machines also shared a preview of Inkling-Small, a lighter-weight model with 12 billion active parameters, which offers strong performance with lower cost and latency.
The startup wants to make customization accessible for more use cases. “Picking the right base model to fine-tune is a qualitative judgment that combines measurable benchmarks with the unique feel of a model that comes from playing with it,” it said in the official statement.
Last week, the startup published a post as a rundown to the announcement, explaining why centrally trained AI that stays fixed underperforms AI that organizations customize themselves, as much expertise lives with the specific people who hold it. Thinking Machine aims to change that by offering customization capabilities. “Even with the best intentions, a model shaped in one place inevitably encodes the values of its owner, not the individual users it serves,” it said.
With a workforce of nearly 200 employees, the launch of Inkling marks the company’s first major public product after about a year and a half of building in stealth mode.
