# Reflection AI runs 1.3 billion sandboxes in its largest RL run

> Original title: We ran ~1.3 billion sandboxes in our largest RL run for Beam - spanning RL training, evaluation, and ablations. With evals and other RL e...

- Source: Reflection
- Published: 2026-10-27T16:36:50.000Z
- Bad Influence: https://badinfluence.news/items/tkqkb813niab5ckuukpmfr7qf
- Original: https://x.com/reflection_ai/status/2115120550230020113

## Summary

We ran ~1.3 billion sandboxes in our largest RL run for Beam - spanning RL training, evaluation, and ablations. With evals and other RL experiments running, the fleet peaked at 883,006 vCPUs (83% of allocatable) across 23 clusters on @googlecloud and @awscloud. We plan to use @OpenEnv (now part of @AgenticAIFdn) for future runs, bringing our learnings to the open ecosystem.

## Post

We ran ~1.3 billion sandboxes in our largest RL run for Beam - spanning RL training, evaluation, and ablations. With evals and other RL experiments running, the fleet peaked at 883,006 vCPUs (83% of allocatable) across 23 clusters on @googlecloud and @awscloud. 

We plan to use @OpenEnv (now part of @AgenticAIFdn) for future runs, bringing our learnings to the open ecosystem.

[Video](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Ftweet_video_thumb%2FHVpq5kZWsAA-ZLj.jpg&mode=thumb&exp=1793404800&sig=c965397fb0fc417a)