Kumo
A pre-trained decision classification model developed by Hopsworks, optimized for low-latency real-time inference scenarios. The model supports on-premise deployment on CPU hardware.
Timeline (last 90 days)
The Hopsworks team benchmarked three pretrained decision models (Kumo, Jev, and Qwen) using a computer game inspired by Subway Surfers.
In this benchmark, Kumo led with an average distance of 2,119 meters over 500 runs.
Kumo and Qwen are local models hosted on Hopsworks and run in the same environment as the game, while Jev is a remote model hosted in another data center.
All Facts (3)
The Hopsworks team benchmarked three pretrained decision models (Kumo, Jev, and Qwen) using a computer game inspired by Subway Surfers.
50%UnverifiedIn this benchmark, Kumo led with an average distance of 2,119 meters over 500 runs.
50%UnverifiedKumo and Qwen are local models hosted on Hopsworks and run in the same environment as the game, while Jev is a remote model hosted in another data center.
50%