General Inference-Side Optimization Evaluation
0Per-Call Latency (ms)Vendor prompt caching
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Vendor prompt cachingDeterministic?=No (cache key opaque), Cross-vendor?=No (vendor-specific), Deployment surface=vendor-managed2026.05 | 0 | — | — | |
| MerlinDeterministic?=Yes (within ISA), Cross-vendor?=Yes, Deployment surface=under 4 MB binary, no third-party deps2026.05 | 1 | — | — |