Perplexity Open-Sources Lily, Its On-Device Inference Engine for Hybrid Compute
Key Info
Perplexity has open-sourced Lily, the local inference engine built for hybrid compute in Perplexity Computer. It is optimized for Qwen3.6-35B-A3B on Apple silicon so on-device processing doesn't bottleneck Computer tasks.
Highlights
- Lily is purpose-built for local inference with Qwen3.6-35B-A3B on Apple silicon.
- It supports Perplexity's hybrid compute approach, keeping on-device execution efficient alongside cloud steps.
- The engine is now available as open source, with more details in the linked post.