Mira Murati's Thinking Machines Lab released its first AI model Wednesday, but the real story isn't the 975-billion-parameter model itself. It's how Murati plans to make money from it.
The startup launched Inkling as an open-weights model, meaning anyone can download, inspect, and modify it for free. Weights are available on Hugging Face. Instead of charging per API call like OpenAI and Anthropic, Thinking Machines will generate revenue through Tinker, a paid platform for fine-tuning the model to specific business tasks.
"The Inkling model" may not be the headline act, Constellation Research analyst Holger Mueller told SiliconANGLE. "Thinking Machines is charging for Tinker, the platform that companies will likely want to use to customize Inkling for their specific use cases."
Inkling is a Mixture-of-Experts Transformer with 975 billion total parameters, though only 41 billion are active for any given prompt. It was trained from scratch on 45 trillion tokens of text, images, audio, and video, developed in less than nine months on Nvidia's GB300 NVL72 systems under a partnership announced in March.
The model natively handles text, images, and audio with a context window of up to one million tokens. It also features adjustable "thinking effort" controls that let developers trade processing speed for accuracy.
Uniquely, Inkling flags its own outputs for uncertainty rather than simply generating confident-sounding hallucinations. On the Artificial Analysis Intelligence Index, Inkling scores 41, making it the top open-weights model from a U.S. lab ahead of Nvidia's Nemotron 3 Ultra at 38. But it still trails leading Chinese open models on text and code tasks, a gap the company acknowledges.
"Inkling is not the strongest overall model available today," the announcement states. The company is betting that customizability matters more than raw benchmark scores.
In a collaboration with Bridgewater Associates, researchers used Tinker to fine-tune an open model with specialized financial data. The resulting lightweight model scored 84.7% on financial reasoning benchmarks, outperforming proprietary alternatives at less than 10% of the cost.
Thinking Machines was founded in February 2025 by former OpenAI executives including Murati, who served as CTO, OpenAI co-founder John Schulman, and former VP Lilian Weng. The startup secured the largest seed round in history, valuing it at $12 billion out of the gate.
Futurum Group analyst Mitch Ashely told the Wall Street Journal that Chinese AI firms have dominated the open-weight ecosystem for the last year, and Inkling is the first Western alternative to those systems. "It gives Western enterprises a credible alternative positioned on customization economics," he said, "shifting spend from per-token API pricing to infrastructure the enterprise controls."
Thinking Machines is also previewing Inkling-Small, a compact version with 276 billion total parameters and 12 billion active. On GPQA Diamond, it scores 88.3%, outperforming its larger sibling at 87.2%.
Full weights will be published once testing is complete.
