---
title: flux-text-encoders
canonical_url: "https://www.modelscope.cn/models/unsloth/flux-text-encoders"
md_url: "https://www.modelscope.cn/models/unsloth/flux-text-encoders.md"
repository: unsloth/flux-text-encoders
last_updated: 2026-08-06
license: other
base_model:
  - openai/clip-vit-large-patch14
  - google/t5-v1_1-xxl
base_model_relation: finetune
parameters: 5.0B
tensor_type:
  - F16
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 68
stars: 0
tags:
  - text-encoder
  - diffusion-single-file
  - comfyui
  - flux
---

# flux-text-encoders

> flux-text-encoders - unsloth 在 ModelScope 开源的模型。FLUX.1 text encoders (mirror)

unsloth/flux-text-encoders 是 ModelScope 魔搭社区上的 5.0B 参数机器学习模型，采用 other 许可，基于 openai/clip-vit-large-patch14、google/t5-v1_1-xxl 构建。

- **Repository**: unsloth/flux-text-encoders
- **License**: other
- **Parameters**: 5.0B
- **Base model**: openai/clip-vit-large-patch14, google/t5-v1_1-xxl
- **Tags**: text-encoder, diffusion-single-file, comfyui, flux
- **Downloads**: 68
- **Stars**: 0
- **Last updated**: 2026-08-06

Source: https://www.modelscope.cn/models/unsloth/flux-text-encoders

---

# FLUX.1 text encoders (mirror)

The two text-encoder files FLUX.1 needs to run under
[stable-diffusion.cpp](https://github.com/leejet/stable-diffusion.cpp), mirrored from
[comfyanonymous/flux_text_encoders](https://huggingface.co/comfyanonymous/flux_text_encoders).

Unsloth Studio fetches these on the no-GPU (native sd.cpp) path. It used to fetch them straight
from the community repack; nobody there owes anyone availability, and a rename or a takedown
turns every CPU FLUX.1 load into a 404. So they are mirrored here. **The weights are unmodified:**
byte for byte the files of the same name in the source repo.

| File | Size | Upstream model | Licence |
|---|---|---|---|
| `clip_l.safetensors` | 0.23 GiB | [openai/clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) (text tower) | MIT |
| `t5xxl_fp16.safetensors` | 9.11 GiB | [google/t5-v1_1-xxl](https://huggingface.co/google/t5-v1_1-xxl) (fp16) | Apache-2.0 |

Only these two files are mirrored. The source repo also ships fp8 T5 variants; Studio does not
use them, so they are not here.

## Licence

Two licences, one per file, both reproduced in full in
[`LICENSE`](https://huggingface.co/unsloth/flux-text-encoders/blob/main/LICENSE), with the
attribution in [`NOTICE`](https://huggingface.co/unsloth/flux-text-encoders/blob/main/NOTICE).
Both permit redistribution.

One caveat worth stating rather than glossing: `openai/clip-vit-large-patch14` carries **no
licence declaration at all**. The MIT terms here are the ones OpenAI attached to
[github.com/openai/CLIP](https://github.com/openai/CLIP), the repository that published those
weights. The source repack labels both its files `apache-2.0`, which is not right for CLIP-L, so
this mirror splits the two rather than repeating that.

Nothing here is an official OpenAI, Google or ComfyUI product, and nothing here is endorsed by
them.