Threads @alternativetonet — Dc98RL6mvOy
Perplexity has open-sourced Lily, its Apple Silicon inference engine for Qwen3.6-35B-A3B, with higher benchmarked prefill and decode throughput.
Perplexity has open-sourced Lily, its Apple Silicon inference engine for Qwen3.6-35B-A3B, with higher benchmarked prefill and decode throughput.