> ## Documentation Index
> Fetch the complete documentation index at: https://docs.freesolo.co/llms.txt
> Use this file to discover all available pages before exploring further.
# Flash
> Train, deploy, and chat with custom LoRA models on managed infrastructure, from one CLI.
Flash is Freesolo's managed post-training service. Write a short config, run one
command, and Flash fine-tunes a model on managed infrastructure, then serves the
result behind an OpenAI-compatible endpoint. Nothing to host. Every command that
talks to the platform authenticates with your Freesolo API key.
Install the CLI and go from an empty directory to a deployed model in a few
minutes.
## What you can do
### Fine-tune a model on your own task
Write a TOML config, run one command, and Flash trains a LoRA adapter (a
small set of add-on weights) on top of a [supported base model](/reference/models). Pick the model and task; Flash handles the training infrastructure. See
[Training](/guides/training).
```bash theme={null}
flash train config.toml
```
### Teach a format from examples with SFT
Supervised fine-tuning learns from the prompt/answer pairs in your [dataset](/guides/datasets).
Use it when you already have examples of the output you want.
```toml theme={null}
algorithm = "sft"
```
### Reward the behavior you want with GRPO
Reinforcement learning scores the model's own completions with your
[environment's](/guides/environments/overview) reward, then updates toward
higher-scoring completions (see [How Flash works](/how-flash-works)). Use it when
you can score an output but can't hand-write the perfect one.
```toml theme={null}
algorithm = "grpo"
```
### Distill a stronger model with OPD
On-policy distillation has a managed teacher (GLM 5.2 by default) grade your
model's own completions token by token, then pulls your model toward it. Use it when a bigger
model already does the task and you want a small one to match it — no answers to
write and no reward to design.
```toml theme={null}
algorithm = "opd"
```
### Serve it behind an OpenAI-compatible API
`flash deploy` registers the adapter with managed serving, then
`flash chat` or any OpenAI client can call it with your Freesolo key. See
[Deploy & chat](/guides/deploy-and-chat).
```bash theme={null}
flash deploy
```
### See the cost before you spend
`--cost` prints the pre-flight estimate without submitting. You pay for the
quoted training run cost and for the tokens you serve. Cancelled runs are
repriced to the training steps they actually reached.
```bash theme={null}
flash train config.toml --cost
```
## Where Flash fits
| I want to... | Start here |
| -------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- |
| Fine-tune from examples I already have | [Training: SFT](/guides/training#choose-a-training-algorithm) |
| Improve outputs I can score but can't hand-write | [Training: GRPO](/guides/training#choose-a-training-algorithm) and [Environments](/guides/environments/overview) |
| Match a stronger model without answers or a reward | [Training: OPD](/guides/training#choose-a-training-algorithm) |
| Define the task and reward my model learns on | [Environments](/guides/environments/overview) |
| Call my model from my existing OpenAI code | [Deploy & chat](/guides/deploy-and-chat#use-it-from-your-own-code) |
| Understand how a run works | [How Flash works](/how-flash-works) |
## Next steps
Install the CLI, log in, and ship your first run in a few minutes.
The loop behind a run: base models, environments, algorithms, serving.
Write a config, submit a run, and follow it to completion.
Serve an adapter, then chat with it over an OpenAI-compatible API.