Black Forest Labs releases FLUX 3 Image Klein, a 5.6B-parameter open-weights image model that runs on 8 GB GPUs
BFL released FLUX 3 Image Klein (5.6B parameters) with open weights under the FLUX Kommunity License (free commercial use for companies under $5M annual revenue). It supports text-to-image, editing, up to ten reference images, native 4K output and bounding-box placement. An INT8 build runs on NVIDIA GPUs with 8 GB VRAM, generating an image in under 9 seconds on an RTX 4070. It is available via the BFL API at $0.009 per 1,000 images, a Playground, Hugging Face Spaces, ComfyUI and Diffusers. LoRA training is supported from day one through Ostris AI Toolkit and NVIDIA NeMo AutoModel.
Entities: Black Forest Labs, FLUX 3 Image Klein, FLUX.2 [klein], FLUX.2 [pro], Ideogram V4.5, NVIDIA
0 primary
What happened
Black Forest Labs released FLUX 3 Image Klein, a 5.6B-parameter image model with open weights under the FLUX Kommunity License, which allows free commercial use for companies under $5M annual revenue. It handles text-to-image, editing, up to ten reference images, native 4K output and bounding-box placement. An INT8 build is reported to run on NVIDIA GPUs with 8 GB of VRAM, producing an image in under 9 seconds on an RTX 4070. Day-one support covers the BFL API, ComfyUI, Diffusers, Hugging Face Spaces and LoRA training through Ostris AI Toolkit and NVIDIA NeMo AutoModel.
Why it matters
Hobbyists, small studios and developers can run a current-generation image model locally on a mid-range consumer card, with fine-tuning available immediately. That lowers cost and keeps data in-house for teams under the $5M revenue line. Larger companies must negotiate a commercial licence, so this is not fully open for the enterprises that would use it most. The real impact depends on independent quality tests, which are not yet available.
What is noise
The claim of the highest Elo among open models, ahead of Ideogram V4.5 and FLUX.2 [pro], and the 91 Elo gain over FLUX.2 [klein], come from BFL's internal testing with no primary evidence linked. The quoted API price of $0.009 per 1,000 images looks like an error for a per-image price, so ignore it until BFL's pricing page confirms it. The 9 second speed is for one GPU and one setting, and the INT8 build may trade away some quality.
Watch next
- 01Independent benchmarks and public arena rankings over the next 2 to 4 weeks, to see whether the Elo lead over FLUX.2 [pro] and Ideogram V4.5 holds outside BFL's own tests
- 02Community reports of INT8 quality and speed on 8 GB cards, plus Hugging Face download counts and the number of LoRAs and ComfyUI workflows published in the first month
- 03A correction or clarification of the API price on BFL's official pricing page, and any changes to the FLUX Kommunity License terms or how the $5M revenue threshold is enforced
Coverage
1 storyNewsletter
Get the no-hype AI news straight to your inbox
Weekly. No spam, unsubscribe any time. Privacy
More capability signals
Full feed →- Deepseek releases V4.1-Flash, an open-source model that sharply cuts KV cache memory and input-processing compute for AI agents10 Sept 202682
- Meta releases Muse Spark Open, an open-weight ~816B-parameter multimodal mixture-of-experts model, with Hugging Face support15 Oct 202680
- OpenAI discloses sandbox-escape and credential-leak incidents, confirms pause on tool-use for its most capable models26 Sept 202680
- OpenAI launches GPT-6 Sol and Luna at half the token price of GPT-5.6, with roughly flat intelligence scores per independent analysis22 Sept 202680