FLUX
By Black Forest Labs, founded by the researchers behind Stable Diffusion and latent diffusion itself. FLUX.2 spans a closed frontier tier and Apache-2.0 open [klein] variants.
The picks.FLUX models by Black Forest Labs
FLUX 3 Video
Multimodal video generation with native synchronized audio across styles and modes
FLUX.2 [klein] 9B KV
KV-cache accelerated image generation and editing for real-time multi-reference workflows
FLUX Virtual Try-On
Low-latency virtual try-on for transferring garments onto a person image with strong identity and garment fidelity
FLUX.2 [max]
The latest state-of-the-art model from Black Forest Labs, generating images grounded in live web information.
FLUX Video Upscale
FLUX-powered video upscaling with precise and creative enhancement modes
FLUX.2 [klein] 9B Base
Undistilled foundation model for high-quality image generation and editing
FLUX.2 [klein] 4B Base
Compact undistilled model for efficient image generation and editing
Common questions
FLUX 3 Video, released August 2026 per the live catalog. New FLUX releases join this page automatically when they land on Runware.
Yes — FLUX.2 [klein] 9B KV, FLUX.2 [dev], FLUX.2 [klein] 9B, FLUX.2 [klein] 4B and 8 more run on Runware's own optimized compute (the platform's open-weight tier, billed on compute time). Check each model page for license terms before self-hosting.
Production notes
Every price on this page is the model's published rate from the live Runware catalog, using the cheapest listed configuration unless stated otherwise. Prices vary with resolution, duration, quality tier, or token volume, so check the pricing table on each model page before estimating unit economics.
The Runware catalog does not publish per-model latency figures, so this page does not quote end-to-end timings. Where a model's own description commits to speed (for example sub-second generation or realtime streaming), that claim is repeated here. For anything else, benchmark the exact models in the Playground with your own payload sizes before committing to an SLA.
Models are addressed by versioned AIR identifiers, so a workflow pinned to specific model versions keeps producing the same behaviour as new versions ship. Adopt upgrades deliberately by re-running your evaluation set against the new version before switching production traffic.
About this collection
Black Forest Labs flagship image family. Strong prompt fidelity and consistency. Klein variants ship open weights under Apache 2.0. This page lists every FLUX model available through the Runware API, published by Black Forest Labs.
Membership is derived from the live Runware catalog: the 26 models on this page are the catalog's current FLUX releases, ordered by the catalog's own editorial weight and release date. Names, descriptions, pricing, capability chips, samples, and guides are all read live from the catalog — nothing here is hand-curated, and new FLUX releases join automatically when they land on Runware.
Everything on this page re-derives from the catalog on every visit. When Black Forest Labs ships a new FLUX model on Runware it appears here with no editorial lag; deprecated versions remain listed with their live status chip so version pins stay auditable.
Model guides.Learn how to use the stack
Audio and speech with FLUX 3
How to compose FLUX 3's synchronized audio: layering ambient and effects, directing music by instrumentation and tempo, directing speech delivery, and rendering multilingual dialogue with clean lip-sync.
Read the guide →Keyframes with FLUX 3
How to pin images to specific frame positions in FLUX 3 videos: opening on a source image, storyboarding across positions, ending on a packshot, morphing between two states, and using timestamps for beat-precise timing.
Read the guide →Multi-shot sequences with FLUX 3
How to build multi-shot video sequences inside one FLUX 3 generation: HARD CUT syntax, shot contrast, time compression, threading an audio bed across the cuts, and pacing shot rhythm.
Read the guide →Prompting FLUX 3
How to prompt FLUX 3 for text-to-video with synchronized audio: request shape, prompt rewriting, camera language, dimensions, style diversity, and the draft-mode iteration workflow.
Read the guide →Video continuation with FLUX 3
How to continue a source clip from its final frames with FLUX 3's video input: single-clip continuation, chaining across multiple generations, deliberate source design, and recovering shots that didn't land.
Read the guide →Virtual try-on
How to dress a person in any garment from a reference image with FLUX VTO. The call takes one person photo, one garment photo, and a short prompt.
Read the guide →Removing objects with FLUX Erase
How to remove objects from images with FLUX Erase, Black Forest Labs's prompt-less mask-driven removal model. What you paint is what disappears.
Read the guide →Extending images with FLUX Outpainting
How to extend an image past its original frame with FLUX Outpainting. No prompt, no mask, just more canvas around what is already there.
Read the guide →Precise and creative enhancement
How to use FLUX Video Upscale's settings.creativity: precise, source-faithful sharpening for faces and brands, or creative detail restoration for scenery, textures, and crowds.
Read the guide →Steering the enhancement with a prompt
How to steer FLUX Video Upscale with an optional positivePrompt: name the materials and textures you want rebuilt so the enhancement favours the detail that matters.
Read the guide →Upscaling video with FLUX Video Upscale
How to upscale video with FLUX Video Upscale: the async request, choosing an upscale factor, the input size and length limits, and keeping motion and the original audio.
Read the guide →Related collections.More ways to build
Browse all collections→Wan
Alibaba's open multimodal family. Strong video and image generation, with open-weights variants for self-hosting.
Open →IMGKling
High-end video generation with strong motion coherence. Production-grade 4K variants, image-to-video pipelines, and avatar generation.
Open →IMGNano Banana
Google's realtime image generation family. Optimised for low-latency interactive apps and high-volume batch inference at production scale.
Open →IMGGPT Image
OpenAI's image generation family. Strong text rendering, instruction following, and a consistent style across short and long prompts.
Open →