What is FLUX 3? The family map: image, video, audio, action
BFL's docs describe FLUX 3 as one family covering image, video with synchronized audio, audio and action. Which pieces have open weights and which do not.

FLUX 3 is Black Forest Labs' unified model family. Its documentation lists four modalities: image, video with synchronized audio, audio, and action. The open weights the docs point to are the earlier FLUX.2 [klein] and [dev] models on Hugging Face, while FLUX 3 Action is a separate open-weights release.
The family at a glance
The BFL documentation describes FLUX 3 as a single family rather than a set of unrelated products, and lists an MCP integration alongside the API. A September 23 post introduces FLUX 3 Action as a 7B open-weights World Action Model.
| Piece | What the sources say | Open weights stated? |
|---|---|---|
| FLUX 3 image | Part of the unified family | Not stated |
| FLUX 3 video | Video with synchronized audio | Not stated |
| FLUX 3 audio | Part of the unified family | Not stated |
| FLUX 3 Action | 7B World Action Model | Yes, open weights |
| FLUX.2 [klein] and [dev] | Earlier generation | Yes, on Hugging Face |
How to read the open-weights column
Where the sources say nothing, the table says not stated. That is not a claim that weights are closed; it means the pages read on October 3, 2026 do not list them. Before you plan a local deployment of any FLUX 3 piece, look for the weights file and the license on the model's own page rather than inferring from the family name.
The Action model is the clearest example of how the family differs from the image models. BFL reports 42.2 percent on RoboLab-120 against 36.8 percent for Cosmos 3 Nano, which is a robotics benchmark. It is not an image or video generator, so it does not help with creative work.
Which piece matters for video makers
For image and video production, the relevant modalities are image and video with audio. A hosted API is the usual way to reach them. On Sume, GET /v1/images/models lists the image models you can call, and each entry publishes its parameters, so you can see whether a given FLUX model is available and which inputs it accepts without relying on a vendor's family page.
A model's presence on a vendor's documentation page and its presence in a particular catalog are separate facts. Always check the catalog you will call.
curl "https://api.sume.com/v1/images/models" \
-H "Authorization: Bearer $SUME_API_KEY"Takeaway
Treat FLUX 3 as a family name, not a license. Confirm modality, availability and weights for the specific model you plan to use.
Sources
Related posts
More in Models
- Where can I use MAI-Voice-2.1 today, and what Sume offers instead
Microsoft lists its Playground, Copilot Audio Expressions and Foundry; the launch post adds OpenRouter and Vercel. What Sume's text to speech gives you instead.
- Check Sume image-model ratios in code before a channel render
Before rendering one product photo for four marketplaces, read GET /v1/images/models and pick a model whose aspect_ratio values cover every channel you sell on.
- Which Kling 3.0, Seedance 2.0 and Wan 3.0 ratios count as a Short?
YouTube makes a Short from a square or vertical video up to 3 minutes. Of Sume's ratios, 9:16, 3:4 and 1:1 qualify; 16:9, 4:3 and 21:9 do not.
- Which Sume image models cannot edit a photo? Five reject references
Imagen 4 Fast and Ultra, Recraft V4, Qwen Image Max and Soul are text-to-image only on Sume. List them from the catalog in 10 lines before you wire up an edit.
Written by Sume