SiliconBench: Speed, Memory, and Fidelity for LLM Serving on Unified-Memory Desktops

SiliconBench, a new benchmark, evaluates nine Apple Silicon serving engines for local LLM serving on unified-memory desktops across three lenses: speed, memory, and fidelity. The benchmark's authors state that concurrent local LLM serving must preserve memory headroom and output fidelity, which speed-only rankings overlook. SiliconBench covers chat and agent scenarios.

Concurrent local LLM serving on unified-memory desktops must preserve memory headroom and output fidelity, which speed-only rankings overlook. We introduce SiliconBench, which evaluates nine Apple Silicon serving engines through three lenses: speed, memory, and fidelity. We evaluate chat and agent s