atlas.bench is a professional-grade, multi-command benchmarking tool that brings statistical rigor to the terminal. Whether you are comparing two different languages, optimizing a hot loop in your code, or auditing system performance, Atlas Bench provides the data you need in a beautiful, synchronized TUI dashboard.

• 🏎️ Side-by-Side Comparison: Instantly see which command is the "Winner" with high-contrast TUI indicators.
• 🔬 Statistical Rigor: Calculates Mean, Median, Min/Max, and Standard Deviation to filter out noise.
• 🔥 Intelligent Warm-up: Automatically performs unmeasured "warm-up" runs to ensure disk caches and OS scheduling are primed.
• 📊 Stability Metrics: Standard Deviation tracking helps you identify "jittery" processes that are fast but inconsistent.
• 🎨 High-Visibility TUI: Built with the Atlas "Onyx & Gold" aesthetic for maximum readability in any terminal environment.
• 📦 Zero Dependencies: Compiles to a single, static binary for zero-friction deployment.

This project uses the Atlas gobake orchestration tool for cross-platform builds.

git clone https://github.com/fezcode/atlas.bench
cd atlas.bench
gobake build

Binaries for your OS and architecture will be located in the build/ directory.

go build -o atlas.bench .

Compare the execution time of two different runtimes:

atlas.bench "node app.js" "bun app.js"

Run a deep audit with 100 iterations and 10 warm-up cycles to eliminate outliers:

atlas.bench --runs 100 --warmup 10 "python3 process.py" "go run main.go"

Atlas Bench doesn't just look at the total time; it analyzes the distribution:

• Mean (Average): The standard "middle ground" of all runs.
• Median: The middle value, which is more resistant to "spiky" outliers than the mean.
• StdDev (Standard Deviation): Measures how much the results vary. A high StdDev suggests that the command's performance is being affected by background tasks or inconsistent system resources.
• The "Winner": The command with the lowest Mean execution time across all timed runs.

• Local-First: No data ever leaves your machine. Configuration and results stay in your terminal.
• Concurrency: Benchmarks are run sequentially to avoid resource contention (CPU/IO interference), ensuring the most accurate measurement possible.
• Engine: Built using bubbletea for the state machine and lipgloss for styling. Orchestrated by gobake.

MIT License - Copyright (c) 2026 FezCode.