Get started in minutes
Install the CLI and go from an empty directory to a deployed model in a few
minutes.
These guides describe the current Flash training and serving contracts. Check
your CLI version and the live catalog for the deployment you use. See the
latest changes.
What you can do
Fine-tune a model on your own task
Choose a supported model, prepare your task, and submit a TOML config. See Training.Choose a training approach
One line of config selects the algorithm. Use SFT when you already have example answers, GRPO when you can score an output but can’t hand-write the perfect one, and custom when you need your own loss or training loop. Custom algorithms run locally against a managed GPU. See Training.Serve it behind an OpenAI-compatible API
flash models deploy registers the adapter with managed serving, then
flash models chat or any OpenAI client can call it with your Freesolo key. See
Deploy & chat.
Or serve it in your own cloud account
Useflash serve deploy for serving in your own Modal account and pay
Modal directly. See
Serve in your own account.
See the cost before you spend
--cost returns a pre-flight estimate without starting paid training or
allocating a training GPU. Successful SFT and GRPO runs bill at the accepted quote,
cancellations are prorated and capped at that quote, and serving is billed per
token. Custom runs bill measured session time up to their accepted cap.
See Cost and billing.
Next steps
Quickstart
Install the CLI, log in, and ship your first run in a few minutes.
Choose a training approach
Pick a model and training approach for your task.
Training
Write a config, submit a run, and follow it to completion.
Deploy & chat
Serve an adapter, then chat with it over an OpenAI-compatible API.