Abundant builds RL rollout architecture with Modal Sandboxes
Understand this story
Modal worked with the Abundant team to design and build a reinforcement learning rollout architecture based on Modal Sandboxes, with a focus on avoiding Abundant paying high costs for idle compute. Modal says the architecture was completed by the Abundant team and took about two weeks in total. As of November 2026, Abundant runs more than 2 million Sandboxes per month, supporting its more than 75,000 rollout jobs.
AI overview based on source reports · Updated 14 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Modal BlogHow Abundant designed efficient rollout architecture with Modal Sandboxes
Abundant builds RL environments for frontier labs and has created over 75,000 rollouts, each running in a Modal Sandbox, with more than two million Sandboxes run per month as of November 2026. Working with Modal's team, Abundant placed a rollout monitor inside each Sandbox that sends a heartbeat every 30 seconds, and used Modal's CPU and memory bursting to lower resource requests and reduce idle compute costs. The architecture took about two weeks to implement and powered over 1,300 hours of agent-research time across nearly 500 agent trials for Abundant's Frontier RSI benchmark.
Activity for this story
- Current comparable activity
- 4
- Peak comparable activity
- 5Nov 20 06:00
- Change in the past 24 hours
- –
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.