# BitDance-14B

> BitDance-14B is an open autoregressive text-to-image model with 14 billion parameters, built on Qwen3-14B by researchers including ByteDance and CUHK and released on 17 February 2026. Instead of diffusing a whole image, it writes the picture as binary visual tokens and predicts up to 64 tokens per step, which the paper says makes it over 30× faster than earlier autoregressive models at 1024×1024. The weights are Apache 2.0. On Kenerate AI it runs in the browser for 15 credits per image.

URL: https://kenerateai.com/models/shallowdream204/bitdance-14b
Publisher: Kenerate AI (https://kenerateai.com)
Last updated: 20 October 2026

## Key facts
- Official name: BitDance-14B (64x and 16x)
- Hugging Face ID: shallowdream204/BitDance-14B-64x
- Maker: BitDance team (ByteDance, CUHK researchers)
- Released: 17 February 2026
- Type: Autoregressive · binary tokens
- Size: 14B · from Qwen3-14B
- Licence: Apache 2.0
- On Kenerate: 15 credits per image
- Inputs: Text prompt only
- Sizes on Kenerate: 8, up to 1536×1536

## Steps
1. Open the model — Go to /app/image and pick BitDance 14B (the button on this page selects it for you). Sign in free to get started.
2. Write the prompt — Describe the subject, setting, style, light and camera in plain English or Chinese sentences.
3. Pick a size and seed — Choose one of 8 sizes — square 1024 or 1536, portrait, landscape, 16:9 or 9:16 — and fix the seed to compare prompt changes.
4. Generate and compare — Each image costs 15 credits. Run the same prompt on another open model, such as Flux 2 Klein Base 9B or ERNIE Image, to see which suits your style.

## In depth

### How BitDance makes an image
Most image models are diffusion models. BitDance is autoregressive: it treats the picture as a sequence, the way a language model treats text [1]. Three pieces make that practical at 14B scale [1][2]:
- Binary tokens — each image patch becomes a binary code instead of an index into a fixed codebook.
- A binary diffusion head — samples those binary tokens in continuous space instead of choosing from a softmax.
- Next-patch diffusion — predicts 16 or 64 tokens per step instead of one, the source of the 30× speed-up the paper reports.

### Running BitDance locally vs online
Locally you need a CUDA GPU, Python 3.11 and flash-attention 2.8.2; the model card gives no VRAM figure [2]. Gigazine's demo made a 1024×1024 image in about 30 seconds [4]. On Kenerate nothing is installed: pick BitDance 14B, choose one of 8 sizes and pay 15 credits per image.

### Prompting BitDance
Write in English or Chinese [2][4]. Describe the subject first, then setting, style, light and camera, in full sentences — BitDance reads the prompt with a Qwen3-14B backbone. Keep the seed fixed while you change one thing at a time.

References:
[1] BitDance: Scaling Autoregressive Generative Models with Binary Tokens — arXiv 2602.14041: https://arxiv.org/abs/2602.14041
[2] shallowdream204/BitDance-14B-64x — Hugging Face model card: https://huggingface.co/shallowdream204/BitDance-14B-64x
[3] shallowdream204/BitDance — GitHub: https://github.com/shallowdream204/BitDance
[4] BitDance autoregressive image generation — Gigazine: https://gigazine.net/gsc_news/en/20260218-bitdance-autoregressive-generation-ai

## FAQ
Q: What is BitDance?
A: BitDance is an open-source autoregressive image generator described in the paper “BitDance: Scaling Autoregressive Generative Models with Binary Tokens” (arXiv 2602.14041). Instead of predicting an index into a codebook, it predicts binary visual tokens using a binary diffusion head, and its next-patch diffusion decodes several tokens in parallel. The text-to-image version, BitDance-14B, has 14 billion parameters and is finetuned from Qwen3-14B.

Q: BitDance vs diffusion models — what's the difference?
A: Diffusion models such as FLUX.2 or ERNIE-Image start from noise and refine the whole image over many steps. BitDance is autoregressive like a language model: it writes the image as a sequence of binary tokens, 16 or 64 at a time. The paper reports that this parallel decoding is over 30 times faster than earlier autoregressive models at 1024×1024. On Kenerate you can run both kinds on the same prompt.

Q: Is BitDance on Hugging Face?
A: Yes. The weights are published by the Hugging Face user shallowdream204: BitDance-14B-64x and BitDance-14B-16x for text to image, plus the BitDance tokenizer and ImageNet research models. The code is at github.com/shallowdream204/BitDance.

Q: BitDance-14B-64x vs 16x — which one?
A: Both are 14B text-to-image models. The 64x version predicts 64 tokens per step at 1024 px and was distilled from the 16x model for about three times faster inference with similar quality, per the GitHub README. The 16x version predicts 16 tokens per step and supports 512 and 1024 px.

Q: Who made BitDance?
A: A research team that includes ByteDance and the Chinese University of Hong Kong, according to Gigazine's report on the release. The paper lists 11 authors, and the GitHub project says the work was accepted to NeurIPS 2026.

Q: What licence does BitDance use?
A: Apache 2.0, according to the Hugging Face model card and the GitHub repo, so self-hosted copies can be used commercially under that licence's terms. Images you generate on Kenerate follow Kenerate's terms of service.

Q: Can I use BitDance 14B online without a GPU?
A: Yes. Kenerate AI runs it in the browser as BitDance 14B, with no download, Python 3.11 or flash-attention install. Pick it in /app/image, write a prompt and choose one of 8 sizes up to 1536×1536. It costs 15 credits per image.

Q: Is BitDance 14B free?
A: The weights are free under Apache 2.0, but you need your own CUDA GPU to run them. On Kenerate BitDance 14B is a paid model at 15 credits per image, from one-time credit packs with no subscription. New accounts get 25 credits on sign-up, and the free plan includes Qwen Image 3.0 and Flux 2 Dev.

Q: Can BitDance edit images?
A: Not on Kenerate: BitDance 14B is text to image only, with no reference upload. To edit a picture, use an edit model such as Flux 2 Klein Base 9B (up to 3 references) or Qwen Image 3.0.

Q: What settings does the BitDance model card use?
A: The example script on the Hugging Face card generates at 1024×1024 with 50 sampling steps and a guidance scale of 7.5; 25 steps is faster but may slightly reduce quality. On Kenerate those settings are handled for you — you choose the prompt, size and seed.

## Related pages
- Open source AI image generator: https://kenerateai.com/open-source-ai-image-generator
- ERNIE-Image: https://kenerateai.com/models/baidu/ernie-image
- FLUX.2 [klein] 9B Base: https://kenerateai.com/models/black-forest-labs/flux-2-klein-base-9b
- Z Image: https://kenerateai.com/model/z-image
- LongCat-Image: https://kenerateai.com/models/meituan-longcat/longcat-image
- All models: https://kenerateai.com/model

## Sources
- BitDance: Scaling Autoregressive Generative Models with Binary Tokens — arXiv 2602.14041: https://arxiv.org/abs/2602.14041
- shallowdream204/BitDance-14B-64x — Hugging Face model card: https://huggingface.co/shallowdream204/BitDance-14B-64x
- shallowdream204/BitDance — GitHub: https://github.com/shallowdream204/BitDance
- BitDance autoregressive image generation — Gigazine: https://gigazine.net/gsc_news/en/20260218-bitdance-autoregressive-generation-ai