AI Models We Use

The specific third-party AI models behind every generation on Flux Video, who builds them, and how we access them.

What powers Flux Video

Flux Video does not train its own generative models. Every generation you run is carried out by a third-party foundation model built by Alibaba, ByteDance, Google, Google DeepMind, Kuaishou, MiniMax, OpenAI and Runway.

The tables below name the exact models available on Flux Video today. The model used for a given generation is either the one you select in the generator, or — where Flux Video selects one for you — the model named in the "Underlying model" column. Where our name for a tool differs from the model behind it, that column names the model.

Image models

ModelBuilt byUnderlying modelAvailability
WAN 2.7 ImageAlibabaAll plans
Nano Banana 2 LiteGoogleGoogle nano-banana-2-lite (Gemini flash image tier).All plans
GPT Images 2OpenAIOpenAI GPT Image 2.All plans

Video models

ModelBuilt byUnderlying modelAvailability
Wan 2.6 FlashAlibabaAll plans
WAN 2.7AlibabaAll plans
Seedance V1 Pro FastByteDanceAll plans
Gemini Omni VideoGoogleAll plans
Veo 3.1 FastGoogle DeepMindPaid plan
Kling V3KuaishouAll plans
Hailuo V2MiniMaxAll plans
Runway Gen4 TurboRunwayPaid plan

How we access these models

We reach these models through their providers' own APIs and through commercial inference aggregators, including KIE.ai, fal.ai, Replicate, Cloudflare Workers AI, and Google's Gemini API. The aggregator is a delivery route, not the model author: the "Built by" column above names the organisation that built and trains each model.

Your prompts and uploads are sent to the provider of the model you are using, for the sole purpose of producing your output. We do not use your content to train models, and we do not sell it. See our Privacy Policy for how your data is handled, and our Terms of Service for the rules that apply to what you generate.

Changes to this list

Model availability changes as providers release, update, and retire models. This page is generated from the service's own model configuration, so it reflects what is actually running rather than a separately maintained list. If a model you used previously is no longer listed, it is no longer available on Flux Video.

Questions about a specific model can be sent to our support team, listed on the Support page.