Play the video in the post belowPost on X
- Views
- 60,644
- Likes
- 655
- Reposts
- 27
Read from X on . Counts change daily.
Cris Lenta’s team benchmarked Jev by running 500 real-time agents in parallel inside a 3D environment, shown in a video in the post.
They call the results preliminary: a 500 millisecond average latency and 35 API calls per second, with a naive implementation and no optimization work. Lenta’s takeaway is that model latency is no longer the bottleneck. The post does not describe what each agent asks Jev or how the environment is passed in.



