Local LLM Test Drive: Benchmark Before You Buy Hardware
Experience real local LLM speeds online before purchasing hardware
- Problem
- Someone on Hacker News launched Llmspeed to let users feel how fast a local LLM runs before buying hardware. This reflects that users lack intuitive performance references when choosing hardware for local large models and often buy mismatched configurations.
- Solution
- A website where users select models and hardware configurations (or upload their own benchmark data) to see key metrics like token generation speed and memory usage. It allows comparison across combinations, with all data crowdsourced from the community.
- Users
- Local LLM enthusiasts, developers, and small studios planning to buy hardware to run open-source models.
- MVP
- Preload common model and hardware combinations; display speed/memory comparison charts; allow users to submit their own benchmark results; filter and sort by model or hardware.
- Revenue
- Free browsing; paid access to detailed benchmark reports or affiliate links for hardware recommendations.
- Why now
- Local LLM hardware options are diverse and information is scattered. Crowdsourced community data can quickly create value, and an indie developer can launch at low cost.