Thinking Machines partners with Bridgewater to build AI model that cuts errors by nearly 30%

When the world’s largest hedge fund decides its analysts are spending too much time on document busywork, it doesn’t just buy a ChatGPT subscription. Bridgewater Associates teamed up with Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, to build a custom fine-tuned model that reduces errors by 29.8% compared to the best available frontier models.

The results, published June 30 by Bridgewater’s AIA Labs and Thinking Machines Lab, show the specialized model hitting 84.7% average accuracy across six information-filtering tasks. Leading models like GPT, Claude, and Gemini variants, even when juiced with expert prompt engineering, were stuck in the mid-70s.

How they built it

The model was constructed on the Qwen3-235B base and trained using Thinking Machines’ proprietary Tinker platform. Two training techniques did most of the heavy lifting: interleaved batching delivered a 12.1% accuracy boost, while on-policy distillation added another 3.1%.

The Tinker API handled the infrastructure side, letting the team iterate rapidly without managing GPU clusters directly.

Inference costs dropped by a factor of 13.8x per task compared to frontier models.

What the model actually does

The six tasks the model handles fall into the category of document triage. Relevancy classification, truncation, and labeling are the core functions — the information-filtering steps that happen before the actual analysis begins.

The key ingredient that made this work wasn’t just clever engineering. It was Bridgewater’s proprietary expert-labeled data. General-purpose models are trained on internet text. When you fine-tune on thousands of examples labeled by domain experts who understand exactly what matters in financial documents, the model learns distinctions that no amount of prompt engineering can teach a generic system.

What this means for investors

Bridgewater manages roughly $100B in assets. For the broader AI industry, this partnership validates Thinking Machines Lab’s approach. Murati left OpenAI in late 2024 and launched the startup with a thesis that the next wave of AI value creation would come from customization, not just scaling.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.