> ## 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.

# Supported models

> The base models Flash can fine-tune and serve, with sizes, reasoning, and token prices.

Flash trains a [**LoRA adapter**](/how-flash-works) on a curated catalog of base
models. Set the base model with one line in your config, then list the live
catalog from the CLI:

```bash theme={null}
flash models
```

```toml theme={null}
model = "Qwen/Qwen3.5-4B"   # required — see `flash models` for the options
```

## Catalog

Every catalog model supports [**SFT, GRPO, and OPD**](/guides/training#choose-a-training-algorithm)
and hybrid reasoning ("thinking").

| Model (`model =`)      | Algorithms     | Reasoning |
| ---------------------- | -------------- | --------- |
| `Qwen/Qwen3.5-0.8B`    | SFT, GRPO, OPD | Hybrid    |
| `openbmb/MiniCPM5-1B`  | SFT, GRPO, OPD | Hybrid    |
| `Qwen/Qwen3.5-2B`      | SFT, GRPO, OPD | Hybrid    |
| `Qwen/Qwen3.5-4B`      | SFT, GRPO, OPD | Hybrid    |
| `Qwen/Qwen3.5-9B`      | SFT, GRPO, OPD | Hybrid    |
| `Qwen/Qwen3.6-27B`     | SFT, GRPO, OPD | Hybrid    |
| `Qwen/Qwen3.6-35B-A3B` | SFT, GRPO, OPD | Hybrid    |

## Context window and LoRA rank

Each base model is served at a fixed **context window** (`max_model_len`) and
accepts LoRA adapters up to a per-model **max rank**. `flash train` enforces both
at submit: a config whose `max_context_tokens` exceeds the served context, whose
GRPO/OPD prompt plus `max_completion_tokens` cannot fit, or whose `lora_rank`
exceeds the cap, is rejected.

| Model                  | Context window | Max LoRA rank |
| ---------------------- | -------------: | ------------: |
| `Qwen/Qwen3.5-0.8B`    |          32768 |           128 |
| `openbmb/MiniCPM5-1B`  |          32768 |           128 |
| `Qwen/Qwen3.5-2B`      |          32768 |           128 |
| `Qwen/Qwen3.5-4B`      |          32768 |            64 |
| `Qwen/Qwen3.5-9B`      |          32768 |            64 |
| `Qwen/Qwen3.6-27B`     |          32768 |            64 |
| `Qwen/Qwen3.6-35B-A3B` |           4096 |            64 |

## Serving capacity

Managed serving keeps one engine per base-model tier and loads trained adapters
as LoRA slots on that tier. The smaller catalog models and the 4B serve on Modal
L4 instances; the 9B and 27B serve on an H100 tier and the 35B on an H200 tier.

| Model                  | Serving GPU | Served context | Hot LoRA slots |
| ---------------------- | ----------- | -------------: | -------------: |
| `Qwen/Qwen3.5-0.8B`    | L4          |          32768 |             16 |
| `openbmb/MiniCPM5-1B`  | L4          |          32768 |             16 |
| `Qwen/Qwen3.5-2B`      | L4          |          32768 |             16 |
| `Qwen/Qwen3.5-4B`      | L4          |          32768 |             16 |
| `Qwen/Qwen3.5-9B`      | H100        |          32768 |             16 |
| `Qwen/Qwen3.6-27B`     | H100        |          32768 |             16 |
| `Qwen/Qwen3.6-35B-A3B` | H200        |           4096 |              6 |

## Serving prices

Serving is billed per token after deployment. Prompt and completion tokens have
separate per-model rates, and cached prompt tokens use the model's cached-input
rate. The prices below are **per 1M tokens**; your [Freesolo billing dashboard](/platform#billing)
shows the charges you accrue.

| Model                  | Prompt / 1M | Completion / 1M | Cached prompt / 1M |
| ---------------------- | ----------: | --------------: | -----------------: |
| `Qwen/Qwen3.5-0.8B`    |     \$0.012 |         \$0.060 |           \$0.0024 |
| `openbmb/MiniCPM5-1B`  |     \$0.012 |         \$0.060 |           \$0.0024 |
| `Qwen/Qwen3.5-2B`      |     \$0.024 |         \$0.120 |           \$0.0048 |
| `Qwen/Qwen3.5-4B`      |     \$0.036 |         \$0.180 |           \$0.0072 |
| `Qwen/Qwen3.5-9B`      |     \$0.137 |         \$0.228 |           \$0.0276 |
| `Qwen/Qwen3.6-27B`     |     \$0.510 |         \$3.666 |           \$0.1680 |
| `Qwen/Qwen3.6-35B-A3B` |     \$0.238 |         \$1.518 |           \$0.0792 |

The **Cached prompt** column is the rate for prompt tokens served from the
automatic **prefix cache**. See [Billing](/guides/deploy-and-chat#billing) for how
prefix caching works and when it applies.

## Reasoning ("thinking") models

Every catalog model is **hybrid-reasoning**: it runs with or without an explicit
reasoning step. Turn reasoning on with one line in your config:

```toml theme={null}
thinking = true
```

See the [`thinking` field](/reference/configuration#top-level) in the
configuration reference.

## Choosing a model

<CardGroup cols={2}>
  <Card title="Start small" icon="feather">
    Use `Qwen/Qwen3.5-0.8B` or `2B` to validate your setup and data
    cheaply. Get a run working before you scale.
  </Card>

  <Card title="Scale up for quality" icon="arrow-up-right-dots">
    Move to `Qwen/Qwen3.5-4B` or `9B` once the task is wired up
    and you want stronger results. Larger models cost more per run.
  </Card>
</CardGroup>

<Card title="Pick a model, then train" icon="dumbbell" href="/guides/training#pick-a-base-model">
  Set `model` in your config and submit a run.
</Card>