← The Polylog AI Intelligence Brief
Morning Edition · Saturday, July 4, 2026Published at 6:44 AM EDT · New York
Thinking Machines and Bridgewater Say a Small Custom Model Beats Frontier Models on Finance Tasks
A fine-tuned model built on Bridgewater's expert-labeled data reportedly cut errors by nearly 30% versus GPT, Claude, and Gemini at a fraction of the cost.
Thinking Machines Lab, the startup led by Mira Murati, and the AIA Labs unit of hedge fund Bridgewater Associates published joint research claiming that a custom fine-tuned model, trained on Bridgewater's proprietary expert-labeled data, ou…
Continue the AI Intelligence Brief
Track frontier labs, chips, export controls, model releases, regulation, and AI infrastructure.
- 5 AI intelligence signals a day
- Frontier labs, compute, and chips
- Model releases and AI infrastructure
- Source-grounded analysis with confidence labels
The Global Intelligence Brief stays free.
Part of a tracked trend
Frontier Model Efficiency Gains
Capability per unit of training and inference compute keeps improving, letting newer models match prior frontier performance far more cheaply and gradually loosening the link between raw scale and capability.
Townsquare
Start a discussion in Townsquare.
More from this edition
- Anthropic Ships Claude Sonnet 5, Pushing Near-Flagship Agent Performance Down the Price Curve
- OpenAI Moves GPT-5.6 From Government-Gated Preview Toward Broad Release
- Anthropic Redeploys Fable 5 and Proposes a Cross-Lab Rubric for Scoring Jailbreaks
- Mistral Open-Sources Leanstral 1.5, Saturating a Formal-Math Benchmark
- Meta's Brain2Qwerty v2 Decodes Typed Sentences From Brain Scans at 61% Word Accuracy
- Anthropic Launches Claude Science, Betting on Workflow Over a New Model
- Google Invests $75 Million in A24 for a DeepMind Filmmaking Partnership
- Google Research Ships TabFM, a Zero-Shot Foundation Model for Tables
Comments
0No comments yet.