Snorkel runs millions of agent simulations per month with Modal sandboxes
Understand this story
In a customer case study, Modal says Snorkel uses its GPU and CPU sandboxes and inference services to run agent simulations, reaching millions per month and tens of thousands of sandboxes per hour at peak; Modal is the only vendor that can meet its large-scale GPU sandbox needs. Modal says Snorkel is still in the early adoption stage, with about 10%–15% of sandboxes currently using bursting. Sandboxes are billed at the higher of the resources actually used at a given moment or the resources originally requested. Snorkel also plans to evaluate using Modal for training in the future, to cover the full model improvement loop from training to evaluation.
AI overview based on source reports · Updated 59 min ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Modal BlogHow Snorkel runs realistic agent simulations with Modal
Snorkel uses Modal to power inference for open-weight models it evaluates and to provide GPU and CPU sandboxes for millions of agent simulations per month, peaking at tens of thousands of sandboxes an hour. Modal's VM sandboxes let Snorkel recreate production-like environments, including multi-service systems and GPU-enabled workloads, and its memory bursting feature cut total sandbox costs by around 20% by avoiding over-provisioning. Snorkel also uses Modal's Shared and Dedicated Endpoints to stand up inference for new and fine-tuned models, and plans to evaluate Modal as a training solution.
Activity for this story
Not enough continuous observations to show a trend yet.