/Patent Tracker
OpenAIPatent applicationAI & softwareInterest 6/10

OpenAI patents method to train AI models with less memory

Patent drawing
Patent drawing from the official filing.

OpenAI has filed a patent application for a system that analyzes an AI model to find the most efficient way to train it. The system checks how much memory the model uses and picks a specific training method that fits within that limit while keeping the speed of updates consistent.

Why it mattersThis could help developers train large AI models on hardware with limited memory, making the process more efficient and potentially cheaper.

In the filing’s own words

A method is described. The method includes profiling a model for a learning network having a plurality of layers and associated memory. The layers include weight layers and activation layers. The plurality of weight layers including weights. The method also includes determining, based on the profiling, a training technique for the model on the learning network. The determination of the training technique includes optimizing a latency for at least one training iteration for a capacity of the associated memory. The training iteration(s) include at least one update of the weights for the weight layers.

What exactly they claim (claim 1)

1. A system comprising: a local memory; and at least one processor, coupled to the local memory, configured to: generate a profile of a model for training in a learning network, the learning network comprising compute engines in one or more compute tiles, the profile being generated by determining a descriptor for the model based on at least one training iteration for the model utilizing at least one training technique; select a training technique from a library of candidate training techniques associated with the learning network for training the model, the selection of the training technique being based on the profile of the model, a training iteration latency, and a capacity constraint of the local memory; and train the model by updating a plurality of weights stored in the learning network using the selected training technique.

Facts

Company
OpenAI (OpenAI Opco, LLC)
Official title
TRAINING OPTIMIZATION FOR LOW MEMORY FOOTPRINT
Publication number
US20260320395A1
Application number
19466604
Filed
Feb 2, 2026
Published
Oct 15, 2026
Inventors
1
Claims
20

Classified under

G06N 3/084 Physics & computingG06N 3/048 Physics & computing

More from OpenAI in AI & software

Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.