T5XXL Unchained needs about 9.2 GB of memory at Q8_0 (a 5.5 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 3 quants, from 2.9 GB (Q4_0) to 10.4 GB (F16).
T5XXL Unchained – a 5.2B parameter T5 encoder with extended tokenizer
T5XXL Unchained is a 5.2B parameter encoder model based on the T5-XXL architecture. It uses an extended tokenizer to support Flux uncensoring and other tokenization benefits. The model is released under the apache-2.0 license.
Good at
- Large context tokenization
- Flux uncensoring support
- Extended tokenizer features
Limits
- No vision support
- Requires 8k context for estimates
- No mixture of experts
Best for
- Tokenization tasks
- Flux model preprocessing
- Large language encoding
Who it’s for. Developers needing a large encoder with extended tokenization for Flux or similar pipelines. Suitable for research and production environments that can accommodate its memory footprint.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_0 | 2.9 GB | 3.6 GB | Fits | Get file → | |
| Q8_0 | 5.5 GB | 6.2 GB | Fits | Get file → | |
| F16 | 10.4 GB | 11.1 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_0 (file size 2.92 GB, estimated 6.6 GB memory). For 12 GB VRAM, Q8_0 (5.52 GB, 9.2 GB memory) is suitable. For higher precision, F16 (10.4 GB, 14.1 GB memory) requires more memory.
What GPU do you need for T5XXL Unchained?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: the best quant that fits is Q4_0 (2.9 GB file, about 6.6 GB in use).
- 12 GB VRAM: the best quant that fits is Q8_0 (5.5 GB file, about 9.2 GB in use).
- 16 GB VRAM: the best quant that fits is F16 (10.4 GB file, about 14.1 GB in use).
Which T5XXL Unchained quant should you choose?
Q8_0 (5.5 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
If memory is tight, the smallest file is Q4_0 at 2.9 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about T5XXL Unchained
How much VRAM does T5XXL Unchained need?
About 9.2 GB at Q8_0 with an 8K context window. That is the 5.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run T5XXL Unchained on an 8 GB GPU?
Yes, with Q4_0: the 2.9 GB file needs about 6.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of T5XXL Unchained GGUF?
It depends on the quant: Q4_0 is 2.9 GB, Q8_0 is 5.5 GB, F16 is 10.4 GB.
What context length does T5XXL Unchained support?
The GGUF header lists a trained context length of 512 tokens (512). The KV cache grows with the context you use: about 3.2 GB at 8K and 0.2 GB at 512.
Where can I download T5XXL Unchained GGUF?
The files are on Hugging Face in the Phil2Sat/T5XXL-Unchained-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does T5XXL Unchained use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run T5XXL Unchained on 8 GB VRAM?
Yes, the Q4_0 quantized version uses about 6.6 GB at 8k context, fitting within 8 GB VRAM.
Does this model support vision tasks?
No, the model is an encoder only and does not include vision capabilities.
What license applies to this model?
The model is released under the apache-2.0 license.