SiMa.ai Palette 3.0 runs perception and LLMs concurrently on one chip, claims 3.5x via speculative decoding

krishnan · x · 2026-11-18

krishnan scrutinizes SiMa.ai's Palette 3.0 (released Nov 17): a robot cannot pause perception while its language model thinks. SiMa claims its Modalix chip runs perception, depth, and reasoning models including LLMs/VLMs concurrently in separate Docker containers within tight power and memory limits, with speculative decoding lifting tokens/sec by up to 3.5x. The pairing deserves a hard test.

Related event: krishnan questions SiMa.ai Palette 3.0's single-chip robot AI claims(8 posts)→

Original post →

More from Embodied

Embodied channel →