Understand this story
Modal's official engineering blog published a post on November 6, 2026, describing how Snorkel uses Modal's GPU and CPU sandboxes to run realistic agent simulations. According to the post, Snorkel runs millions of agent simulations each month, peaking at tens of thousands of sandboxes per hour, and uses Modal's inference endpoints to serve the open-weight models being evaluated. The article also states that after adopting Modal's memory bursting, Snorkel no longer needs to over-provision resources, saving about 20% on its total sandbox bill, and currently about 10-15% of sandboxes use this capability.
AI-generated from public reports
Latest development · Report
How Snorkel runs realistic agent simulations with ModalDevelopments
1 confirmed developments
Modal provides inference and sandbox support for Snorkel's Agent simulation
Source:English text is being prepared.
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 6
Modal Official Engineering BlogFirst-party
How Snorkel runs realistic agent simulations with Modal
Using Modal's GPU and CPU sandboxes, Snorkel runs millions of agent simulations per month, peaking at tens of thousands of sandboxes per hour, and uses Modal's inference endpoints to serve the open-weight models being evaluated. After adopting Modal's memory bursting, Snorkel no longer needs to over-provision resources, saving about 20% on its total sandbox bill; currently about 10-15% of sandboxes use this capability.