カオスマップに戻る

Overview

Flux is a family of image generation models from Black Forest Labs (BFL). The current image flagship is FLUX.2. Its successor FLUX 3 was announced on July 23, 2026, but only the video half — FLUX 3 Video — has reached general availability; FLUX 3 Image is not yet generally available. FLUX.2 offers multi-reference conditioning across up to 10 reference images and coherent 4-megapixel generation and editing, placing it at the frontier on both photorealism and prompt following.

Key Features

Multi-reference conditioning

FLUX.2 can pull in up to 10 reference images and preserve the same character, layout and style across generations — a meaningful advantage for comics, ad campaigns and other jobs that need consistency.

Open-weight + commercial endpoints

FLUX.2 ships in multiple variants (Pro / Flex / Dev / Klein). Pro and Flex are proprietary hosted endpoints; Dev is an open-weight download requiring a BFL commercial license; Klein is Apache 2.0 and free to run locally for commercial work.

FLUX 3 pushes into multimodal

FLUX 3 was announced as a multimodal model handling image, video, audio and robot action in one architecture. FLUX 3 Video reached general availability in early August 2026, generating clips up to 20 seconds with native in-sync audio at published rates of $0.06–$0.54 per second. FLUX 3 Image and the open-weight FLUX 3 Dev are still to come.

Plans & Pricing

DeliveryPriceNotes
BFL API1 credit = $0.01Per-image
FLUX.2 [Pro]~$0.03 per MP1024×1024 ≈ $0.030
FLUX.2 KleinFirst 1MP $0.014$0.001 per extra MP
FLUX.2 [Dev]Open-weight DLCommercial needs BFL license

The BFL pricing page currently lists FLUX.2, FLUX Tools and FLUX.1 only — no per-image rate for FLUX 3 Image has been published yet. Flux has no first-party consumer subscription — you consume it pay-as-you-go via API or through hosts like Together.ai, Replicate or OpenRouter.

Typical Use Cases

  • Photoreal ad visuals and product shots
  • Comics, picture books and storyboards that need character consistency
  • Product background swap and variant generation
  • Advanced pipelines combining Dev weights + LoRAs in ComfyUI

Strengths

  • Top-tier photorealism and prompt adherence
  • Up to 10 reference images for consistency
  • Open-weight Dev / Klein variants enable local use
  • Much improved text rendering generation over generation

Weaknesses

  • No polished consumer UI — API / third-party surfaces only
  • Commercial use of the Dev model requires a separate BFL license
  • High-resolution generation can get expensive

Official Info

公式サイト