8080 — Speed Demo

API settings

Base URL

API key

Model

Max tokens

Settings are saved in this browser's localStorage.

Compare models

All providers run in parallel through the proxy using the same prompt and max tokens. Groq, Cerebras and OpenRouter keys come from GROQ_API_KEY, CEREBRAS_API_KEY and OPENROUTER_API_KEY set as server secrets.

Prompt

⌘/Ctrl + Enter

0

output tokens / second

Prefill rate

–

input tokens / sec

Server time

–

total generation

Tokens

–

in / out

Decode rate

–

output tokens / sec

Output speed compared to typical LLM APIs (approximate)

Total response time (lower is better)

Output tokens / second, end to end (higher is better)

Times are measured at the proxy from request sent to full response received, so they include network latency and queueing. Models differ between providers, so output lengths can differ too. Tokens/second divides each provider's own output token count by its total time.