Sleep Time Compute Letta

Sleep Time Compute Letta Sleep time compute is a new way to scale ai capabilities: letting models "think" during downtime. instead of sitting idle between tasks, ai agents can now use their "sleep" time to process information and form new connections by rewriting their memory state. Code and data accompanying the paper sleep time compute: beyond inference scaling at test time from letta and uc berkeley. this repo contains code to reproduce the empirical aime gsm results in the sleep time compute research paper.

Sleep Time Compute Letta We introduce sleep time compute, which allows models to "think" offline about contexts before queries are presented: by anticipating what queries users might ask and pre computing useful quantities, we can significantly reduce the compute requirements at test time. Researchers from letta and uc berkeley have introduced a novel approach called “ sleep time compute,” enabling large language models (llms) to utilize idle periods for pre processing. The sleep time compute system is designed around a novel computational paradigm that splits reasoning processes into two distinct phases. this approach allows for more efficient use of computational resources and improved performance on complex reasoning tasks. Tl;dr: i used letta’s sleep time compute pattern to compare two very different ways an agent spends compute. test time compute is the classic “think under pressure” mode.

Sleep Time Compute Letta The sleep time compute system is designed around a novel computational paradigm that splits reasoning processes into two distinct phases. this approach allows for more efficient use of computational resources and improved performance on complex reasoning tasks. Tl;dr: i used letta’s sleep time compute pattern to compare two very different ways an agent spends compute. test time compute is the classic “think under pressure” mode. In this week's paper read, we’ll dive into a groundbreaking new paper from researchers at letta, introducing sleep time compute: a novel technique that lets models do their heavy lifting offline, well before the user query arrives. The “sleep time compute” concept introduced by letta proposes that agents benefit from scheduled downtime, during which they process chat histories, reorganize context, and essentially “refresh” their internal models. The term was coined in an april 2025 white paper by letta, a berkeley born ai startup spun out of uc berkeley’s sky computing lab, founded by researchers charles packer and sarah wooders. Sleep time compute — letta ai (charles packer, charlie snell, kevin lin) latent space 45.9k subscribers subscribe.