Developers waste days testing different local LLMs without consistent benchmarks.
Create a Docker image that runs standardized tests across models and hardware configurations.
Sell curated benchmark reports and enterprise features for team deployment tracking.
The MVP could test just throughput and latency on common consumer GPUs.
Risk: Rapid model improvements make benchmarks quickly outdated.