FLUX 3 Dev
#33 in Text-to-VideoBlack Forest Labs · v3 · dev · 2× · tolest 25. Juli 2026
FLUX 3 Dev is the planned Open-Weight variant of Black Forest Labs' (BFL) multimodal FLUX 3 model, which generates images, videos, audio, and action predictions from a single neural network. FLUX 3 was unveiled on July 23, 2026, with initially only FLUX 3 Video and FLUX 3 Action available as gated Early Access; FLUX 3 Dev is officially announced but not yet released. According to official and specialized sources, concrete technical specifications such as parameter count, hardware requirements, licensing terms, price, and exact release date for FLUX 3 Dev have not been published to date. According to the announcement, the model will support Fine-Tuning of weights via standard methods (LoRA, Full Fine-Tuning).
Features
| Fine-Tuning | Planned support: LoRA, full fine-tuning and adapter-based approaches on custom data (per announcement, docs still pending) |
| License | Not finalized yet; BFL has indicated a permissive open-weight license for research and commercial use, details pending |
| Platform | Planned as downloadable open-weight release for local/cloud deployment (e.g. Hugging Face), similar to FLUX.2 Dev |
| Price | Not published (no pricing for FLUX 3 Dev disclosed as of July/August 2026) |
| Release Date | Not yet announced; planned for 'later in 2026', after FLUX 3 Video/Image/Action |