A robot can't pause perception while its LLM thinks: SiMa Palette 3.0 single-chip claim under scrutiny

krishnan · x · 2026-11-18

Analyst krishnan challenges SiMa.ai's Palette 3.0 edge inference platform (released Nov 17), which claims to run several models on one chip with faster token output — but a robot cannot pause perception while its language model thinks. In follow-ups he argues speculative decoding only speeds accepted tokens, not perception freshness during long reasoning; vendor throughput claims aren't robot-safety or task-completion results, and shared silicon means vision, depth and reasoning compete for the same compute and memory. He advises buyers to demand demos of what the rest of the machine does during reasoning, notes a dedicated controller with an independent safety path may be better, and calls resource-budget enforcement a product opportunity.

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

Original post →

More from Embodied

Embodied channel →