FLUX.2 [flex]

Text-to-Image • Black Forest Labs • Proxied

FLUX.2 [flex] is Black Forest Labs' fine-grained control variant of FLUX.2 — exposes tunable inference steps, guidance, and prompt upsampling for typography-heavy and production workflows.

Model Info
Terms and License	link ↗
More information	link ↗
Pricing	View pricing in the Cloudflare dashboard ↗

const response = await env.AI.run(
  'black-forest-labs/flux-2-flex',
  {
    prompt:
      "Samsung Galaxy S25 Ultra product advertisement, 'Ultra-strong titanium' headline, close-up of phone edge showing titanium frame, dark gradient background, clean minimalist tech aesthetic",
  },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "black-forest-labs/flux-2-flex",
  "input": {
    "prompt": "Samsung Galaxy S25 Ultra product advertisement, '\''Ultra-strong titanium'\'' headline, close-up of phone edge showing titanium frame, dark gradient background, clean minimalist tech aesthetic"
  }
}'

Output
Raw response

{
  "state": "Completed",
  "result": {
    "image": "https://examples.aig.cloudflare.com/black-forest-labs/flux-2-flex/typography-design.jpeg"
  },
  "gatewayMetadata": {
    "keySource": "Unified"
  }
}

Examples

High Detail Generation — Crank steps and guidance for maximum detail when latency is not the priority

TypeScript
cURL

const response = await env.AI.run(
  'black-forest-labs/flux-2-flex',
  {
    prompt: 'A detailed oil painting portrait of a Renaissance nobleman with intricate lace collar',
    guidance: 7.5,
    steps: 50,
  },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "black-forest-labs/flux-2-flex",
  "input": {
    "prompt": "A detailed oil painting portrait of a Renaissance nobleman with intricate lace collar",
    "guidance": 7.5,
    "steps": 50
  }
}'

Output
Raw response

{
  "state": "Completed",
  "result": {
    "image": "https://examples.aig.cloudflare.com/black-forest-labs/flux-2-flex/high-detail-generation.jpeg"
  },
  "gatewayMetadata": {
    "keySource": "Unified"
  }
}

Fast Draft — Fast draft with prompt upsampling disabled — preserves the literal prompt

TypeScript
cURL

const response = await env.AI.run(
  'black-forest-labs/flux-2-flex',
  { prompt: 'A simple line sketch of a mountain landscape', prompt_upsampling: false, steps: 10 },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "black-forest-labs/flux-2-flex",
  "input": {
    "prompt": "A simple line sketch of a mountain landscape",
    "prompt_upsampling": false,
    "steps": 10
  }
}'

Output
Raw response

{
  "state": "Completed",
  "result": {
    "image": "https://examples.aig.cloudflare.com/black-forest-labs/flux-2-flex/fast-draft.jpeg"
  },
  "gatewayMetadata": {
    "keySource": "Unified"
  }
}

Parameters

Input
Output

guidance

numbermaximum: 10minimum: 1.5Classifier-free guidance scale (1.5–10). Higher values follow the prompt more strictly at the cost of realism.

height

integermaximum: 9007199254740991minimum: 64Height of the generated image in pixels (minimum 64). Omit to let BFL pick.

▶input_images[]

arraymaxItems: 8Up to 8 reference images for editing or multi-image composition. Each entry is an HTTPS URL or a data:image/...;base64,... URI.

output_format

stringenum: jpeg, png, webpOutput image format. Defaults to jpeg.

prompt

stringrequiredText prompt for image generation or editing.

prompt_upsampling

booleanWhether BFL should expand short prompts before generation. Defaults to true on flex.

safety_tolerance

integermaximum: 5minimum: 0Tolerance for input/output moderation. 0 is the strictest, 5 the most permissive. Defaults to 2.

seed

integermaximum: 9007199254740991minimum: -9007199254740991Optional seed for reproducible generation.

steps

integermaximum: 50minimum: 1Number of denoising steps (1–50). Higher steps yield more detail at the cost of latency.

width

integermaximum: 9007199254740991minimum: 64Width of the generated image in pixels (minimum 64). Omit to let BFL pick.

image

stringformat: uriURL to the generated image

API Schemas (Raw)

Input

Output