← Back to feed

Optimizing On-Device Inference for Apple Silicon

AI · Technology · 2026-11-15 19:27

Perplexity has released Lily, a lightweight on-device inference engine optimized for Apple Silicon. By separating prefill and decode phases, the tool significantly boosts performance for Qwen3 models in local environments.

Source: www.perplexity.ai


Auto-generated summary for reference only. RSS