FLUX.2 Pro
Key facts · verified Oct 28, 2026
FLUX.2 Pro is a Black Forest Labs image generation model released on 25 Nov 2025. Pricing: $0.03 for the first output megapixel + $0.015/MP thereafter; reference images $0.015/MP each - API pricing. It ranks #32 on LMArena's text-to-image leaderboard (Elo 1153, read 1 Nov 2026). Source: bfl.ai.
FLUX.2 [pro] is Black Forest Labs' API-only image generation and editing model in the FLUX.2 family, released on November 25, 2025. BFL documents output up to 4MP, up to 8 reference images through the API (10 in the Playground), and JSON-based pose and positioning control. It is a closed model, distinct from the open-weight FLUX.2 [dev].
$0.03 for the first output megapixel + $0.015/MP thereafter; reference images $0.015/MP each - API pricing
Not published
Key Features
Full Specifications
$0.03 for the first output megapixel + $0.015/MP thereafter; reference images $0.015/MP each - API pricing | |
|---|---|
| Identity | |
| Developer | Black Forest Labs |
| Released | Nov 2025 |
| Status | GA |
| Licence | Proprietary |
| Self-hostable | No |
| Cost | |
| Price per image | Text-to-image: $0.03 for the 1st output MP + $0.015/MP thereafter; image editing: from $0.045 (output MP + $0.015/MP per reference image); reference images $0.015/MP each (1 MP = 1024x1024px, rounded up per image) |
| Cheapest plan | No subscription plan - pay-as-you-go via BFL API/Playground credits only (1 credit = $0.01 USD), billed per output megapixel (see Price per image) |
| Free tier | Not offered - no free API or Playground credits found; BFL's own Quick Start guide has new accounts "Add credits to your account...Start with $10-20 to experiment" before generating anything |
| Capacity | |
| Max resolution | Up to 4MP output (e.g. 2048x2048 at 1:1); width/height each must be >=64px (API); images/references exceeding 4MP are resized down to 4MP |
| Aspect ratios | Flexible - any aspect ratio via width/height parameters (not a fixed enum), up to the 4MP output budget; BFL's own prompting guide lists 1:1, 16:9, 9:16, 4:3, 3:2 and 21:9 as common presets |
| Images per request | 1 output image per API call (single "sample" in the response; no batch/count parameter) - can reference up to 8 input images via the API (10 in the Playground) to inform that single output |
| Capability | |
| Inpainting | Not offered as masked/region inpainting on [pro] itself - no mask parameter on the FLUX.2 [pro] API (image editing is prompt + reference-image driven); masked erase/outpaint ship as separate FLUX Tools endpoints built on FLUX.2 [klein] 9B, not on [pro] |
| Image-to-image | Yes - text-and-image-to-image editing via input_image through input_image_8 (up to 8 reference images via the API, 10 in the Playground) |
| Text rendering | Yes - "Production-ready text": complex typography, infographics, memes and UI mockups with legible fine text work reliably in production |
| Style controls | Yes - style transfer via reference images (e.g. "apply the style of image X") and text prompting; no separate numeric style-weight parameter |
| Character consistency | Yes - up to 8 reference images via the API (10 in the Playground) for character/product/style consistency across generations |
| Rights | |
| Commercial rights | Yes - BFL claims no ownership of Output; Developer and End Users may use Outputs for personal or commercial purposes, subject to the Developer Terms and Usage Policy |
| Training-data opt-out | Yes - prompts, uploads and outputs are used for training/improvement by default; opt out by emailing privacy@blackforestlabs.ai with subject "Training Opt Out"; no self-serve toggle |
| Access | |
| API | Yes (BFL API - flux-2-pro / flux-2-pro-preview endpoints) |
| Chat app | Yes (BFL Playground - web UI for [max]/[pro]/[flex], no code required; draws from the same credit pool as the API, not a separate free product) |
| Cloud marketplaces | Microsoft Foundry (Azure) - listed as a "Direct from Azure" model, Generally Available; not found on AWS Bedrock or Google Cloud Vertex AI marketplaces as of this verification |
| Fine-tuning | Not offered on [pro] itself (closed/API-only) - LoRA/full fine-tuning is only available for the open-weight FLUX.2 [klein] models (via community tools) and self-hosted [dev] |
| Measured quality | |
| LMArena Elo | 1153 (checked Nov 2026) |
| LMArena rank | Rank 32 |