# FLUX.2 [pro]

FLUX.2 [pro] is the FLUX.2 generation foundation model. It pairs a rectified-flow transformer with a Mistral-3 24B VLM, supports up to 10 reference images, and outputs at 4 MP. This is Black Forest Labs's architectural leap beyond FLUX.1.

- **Model ID:** `bfl/flux-2-pro`
- **Type:** image
- **Providers:** bfl
- **Context window:** 67,300
- **Maximum output tokens:** 67,300
- **Pricing:** See the provider details below
- **Canonical page:** https://vercel.com/ai-gateway/models/flux-2-pro

## Supported parameters

- **Parameters:** Not reported
- **Operations:** `text-to-image`, `image-to-image`
- **Input modalities:** `text`, `image`
- **Output modalities:** `image`
- **Aspect ratios:** Not reported
- **Resolutions:** Not reported
- **Durations:** Not reported
- **Frame rates:** Not reported
- **Maximum outputs per request:** Not reported
- **Maximum input images:** 9
- **Input image formats:** Not reported
- **Maximum input image size:** Not reported

## Example

```ts
import { experimental_generateImage as generateImage } from 'ai';

const result = await generateImage({
  model: 'bfl/flux-2-pro',
  prompt: 'A red balloon on a wooden table.'
});
```

## About

FLUX.2 [pro] is Black Forest Labs's FLUX.2 generation launch model, released on November 25, 2025. The FLUX.2 architecture departs from FLUX.1 by pairing a rectified-flow transformer with a Mistral-3 24B vision-language model (VLM) as the text conditioning component, retrained with a new VAE. The VLM handles semantic understanding of complex prompts and reference inputs. The flow transformer handles image synthesis. Together they enable FLUX.2's improved reference image handling.

The multi-reference capability accepts up to 10 input images to guide generation simultaneously. In practice, you can supply a product photo, a lighting reference, a composition reference, and a style example together. The model synthesizes all of them coherently into a new image. Exact color matching (producing output that hits specific hex values or matches reference palette samples) directly benefits brand asset production where color accuracy is non-negotiable.

At up to 4 megapixels output resolution, FLUX.2 [pro] is the entry point to the FLUX.2 generation for teams whose workflows need the multi-reference architecture. FLUX.2 Max extends the quality ceiling further. FLUX.2 [pro] is the generation baseline that introduced the paradigm.

## What to consider

For commercial workflows using branded visual assets as references, confirm licensing terms with your provider for multi-reference generation use. Check N/A before you scale traffic.

## When to use

### Best For

**Multi-reference brand asset generation:** Supplying product photos, brand color palettes, and composition references simultaneously produces on-brand imagery without manual post-production color correction.

**Exact color matching workflows:** Fashion, product photography, and graphic design pipelines where output must match specific color values benefit from the precise color fidelity.

**High-resolution commercial production:** At up to 4MP, output is suitable for print materials, large-format advertising, and other applications where pixel density matters.

**FLUX.2 architecture adoption:** Teams moving from FLUX.1 to the FLUX.2 generation use FLUX.2 [pro] as the baseline model, then evaluate FLUX.2 Max for the quality ceiling.

### Consider Alternatives When

**Iterative multi-turn image editing is the workflow:** FLUX Kontext Pro is purpose-built for sequential edits to the same image, a different paradigm from reference-guided generation.

**Interactive or real-time generation is needed:** FLUX.2 Klein is optimized for sub-second generation, a different performance tier.

**The FLUX.1 workflow already works well:** FLUX 1.1 Pro is the recommended production model for FLUX.1-generation pipelines and may not require an architectural upgrade.

## Best for

- **Multi-reference brand asset generation:** Supplying product photos, brand color palettes, and composition references simultaneously produces on-brand imagery without manual post-production color correction
- **Exact color matching workflows:** Fashion, product photography, and graphic design pipelines where output must match specific color values benefit from the precise color fidelity
- **High-resolution commercial production:** At up to 4MP, output is suitable for print materials, large-format advertising, and other applications where pixel density matters
- **FLUX.2 architecture adoption:** Teams moving from FLUX.1 to the FLUX.2 generation use FLUX.2 [pro] as the baseline model, then evaluate FLUX.2 Max for the quality ceiling

## Consider alternatives

- **Iterative multi-turn image editing is the workflow:** FLUX Kontext Pro is purpose-built for sequential edits to the same image, a different paradigm from reference-guided generation
- **Interactive or real-time generation is needed:** FLUX.2 Klein is optimized for sub-second generation, a different performance tier
- **The FLUX.1 workflow already works well:** FLUX 1.1 Pro is the recommended production model for FLUX.1-generation pipelines and may not require an architectural upgrade

## Frequently asked questions

### What makes FLUX.2 architecturally different from FLUX.1?

FLUX.2 pairs a rectified-flow transformer with a Mistral-3 24B vision-language model (VLM) for text conditioning. This replaces the T5/CLIP conditioning used in FLUX.1. The VLM enables richer semantic understanding of complex prompts and reference images.

### How does multi-reference synthesis work with up to 10 input images?

Submit up to 10 images alongside your text prompt. The model synthesizes a new image that incorporates elements (style, color, composition, subjects) from the references according to your prompt instructions.

### What does exact color matching mean in practice?

The model can reproduce specific color values from reference images or color samples with high fidelity. For brand asset workflows where output must match a defined palette, this reduces or eliminates post-production color correction.

### How does FLUX.2 [pro] compare to FLUX.2 Max?

Both are FLUX.2-generation models. FLUX.2 [pro] is the production-tier release. FLUX.2 Max targets the quality ceiling with additional fidelity and typography improvements for the most demanding commercial work.

### What resolution does FLUX.2 [pro] support?

Up to 4 megapixels. Specific dimension combinations depend on the provider. Refer to available providers for available aspect ratios and resolutions.

### What does FLUX.2 [pro] cost per image?

This page lists the current rates. Multiple providers can serve FLUX.2 [pro], so AI Gateway surfaces live pricing rather than a single fixed figure.

### How do I generate images with FLUX.2 [pro] on AI Gateway?

Use the identifier `bfl/flux-2-pro` with `generateImage` from the AI SDK. Submit text prompts and optional reference images. AI Gateway routes the request to available providers automatically.

## Links

- [Model page](https://vercel.com/ai-gateway/models/flux-2-pro)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://bfl.ai/blog/flux-2)
- [Provider pricing](https://bfl.ai/pricing)