AnalysisAI ModelsOctober 16, 2026

MiMo-2.6-Pro runs on Jetson AGX Thor at 0.646 tokens/sec

Read original source →reddit.com

A Reddit user loaded MiMo-2.6-Pro on a 128GB Jetson AGX Thor, reaching 0.646 tokens/sec with roughly 566GB of model data traced layer by layer. The post contrasts the run's step-by-step inspectability with Dario Amodei's remark that closed models can't be seen inside.

1 source

More stories today

Open the live feed