Resources
Everything we publish, and where to find us.
Engineering notes, the book, product updates rendered from the changelog, the open source project, and the people behind it.
All resources
Blog
Kernel work, measured benchmarks and what it takes to run frontier models on hardware you own. Everything reproducible from a commit.
Read the blog ↗Documentation
The Metrale book. Install, recipes, deployment, the control plane and the gates, with an llms.txt for answer engines.
Open the book ↗Product updates
What shipped, what changed and what it does not change, rendered from the repository changelog on every build.
See what shipped →Benchmarks
The concurrency ladder and every gate record, generated from the repository. Reproduce any of them with the command on the page.
See the numbers →Open source
The engine, under AGPL-3.0. One install command, every recipe, the deployment guide and the Discord where the work happens.
Install the engine →Contributors
Everyone who has landed code in the repository, called out by name, with the core team’s roles.
Meet the contributors →Events
Where to meet the team, in person and online.
Find us →Metrale Labs
The research arm. Kernels, compression, speculative decoding, protocols and the day zero model bring ups.
See the research →Verification walkthrough
Reproduce the ladder yourself. Fingerprint, parity, commands and the artifacts behind the vLLM comparison.
Walk through it →Open source
The engine is free. The platform pays for the people who keep it that way.
Install the engine in one command, run any recipe, and bring your machine to the test fleet. The Enterprise Edition adds the control plane, the economics layer and a named engineer.