howfast

A small CLI tool that measures token metrics (prompt tokens, completion tokens, total tokens, and tokens-per-second) for an Ollama model response. It queries an Ollama server, then prints a nicely formatted, colored summary. The actual model text response is hidden by default; pass --with-response to show it.

Installation

cargo install howfast

Features

• Call an Ollama model and collect token metrics.
• See Tokens per second (completion)
• Nicely formatted, colored terminal output (no box/borders).
• Response hidden by default; opt-in with --with-response.
• Uses Tokio runtime (required by ollama-rs).

Requirements

• Rust (1.70+ recommended)
• Cargo
• An Ollama server reachable from your machine (default localhost:11434)

Environment

• OLLAMA_HOST (optional) — host address of the Ollama server (defaults to localhost). Do not include the port; the program always uses port 11434.

Example:

# set before running, if your Ollama server isn't on localhost
export OLLAMA_HOST=192.168.1.100

or

OLLAMA_HOST=192.168.1.100 howfast gemma3:4b "Tell me a joke"