A new contender has entered the bustling artificial intelligence arena! Thinking Machines Lab, an AI firm founded by former OpenAI executives and researchers, has just unveiled its inaugural model, aptly named Inkling. This release marks a significant moment, positioning the startup as a serious player in the increasingly competitive AI landscape.
What makes Inkling particularly noteworthy is its “open-weight” nature. This means the model’s underlying components are freely available for download, allowing researchers and other startups to inspect, modify, and build upon its foundation. This commitment to openness aligns with a broader vision for decentralized AI development.
Meet Inkling: A New Breed of AI Model
Thinking Machines Lab’s official blog post reveals that Inkling was meticulously trained from the ground up. Its unique capability lies in its ability to process and comprehend not just text, but also audio and video input. While the company acknowledges Inkling might not top every popular benchmark, it excels across a diverse range of tasks, demonstrating impressive capabilities in advanced reasoning and coding.
Like many powerful open-weight models, Inkling is a substantial piece of technology, boasting 975 billion parameters. To run effectively, it requires a dedicated cluster of specialized chips, highlighting its advanced computational demands. Interestingly, in a true testament to AI’s evolving role, Inkling was even leveraged to fine-tune and improve itself during its development.
The Power of Open-Weight AI
The decision to release Inkling as an open-weight model is a strategic move, aligning with a philosophy that advocates for broader access to AI technology. Open-source and open-weight models offer distinct advantages over their closed counterparts. They are often significantly cheaper to operate and can be more readily adapted and customized for specific applications.
While some of the top-performing open-weight models currently hail from China, Thinking Machines Lab asserts that Inkling delivers a comparable level of performance. This positions Inkling as a compelling alternative, potentially democratizing access to high-tier AI capabilities and fostering innovation across the industry.
A Glimpse Behind the Curtain: Inkling’s Unique Discovery
During Inkling’s training phase, researchers at Thinking Machines Lab stumbled upon a fascinating, albeit unexpected, phenomenon. Typically, AI models are designed to provide natural language explanations for their complex reasoning processes. However, Inkling, in a surprising turn, opted to streamline this for efficiency.
As a company source, who chose to remain anonymous, explained, “It determined that the grammar was overhead, which is interesting.” While this efficiency was compelling, the team ultimately reinstated natural language reasoning. This decision was crucial for making the model’s intricate decisions more transparent and explainable to human users.
Thinking Machines Lab: A Legacy of Innovation
Thinking Machines Lab was founded in February 2025 by a star-studded roster of former OpenAI leaders. This includes Mira Murati, who served as OpenAI’s CTO and briefly CEO; John Schulman, an OpenAI co-founder instrumental in ChatGPT’s development; and Lilian Weng, a former OpenAI VP who spearheaded safety and robotics initiatives.
This powerhouse team quickly made waves, securing the largest seed funding round in history, which valued the startup at a staggering $12 billion right from the start. Prior to Inkling, the company had already showcased its innovative spirit with releases like Tinker, a model fine-tuning tool, a natural voice interaction tool, and contributions to machine learning research.
While OpenAI’s ChatGPT undeniably ignited the current AI boom, companies like Thinking Machines Lab and Anthropic, led by similarly prominent defectors, are rapidly carving out their own significant niches. Anthropic, for instance, recently filed for an IPO that could value it at over a trillion dollars, with its Claude model gaining widespread popularity, particularly for its coding prowess. The AI race is accelerating, and Thinking Machines Lab is now a formidable contender.
Source: Wired – AI
