---
title: ltx-2.5-mlx-ditq8
canonical_url: "https://www.modelscope.cn/models/mlx-community/ltx-2.5-mlx-ditq8"
md_url: "https://www.modelscope.cn/models/mlx-community/ltx-2.5-mlx-ditq8.md"
repository: mlx-community/ltx-2.5-mlx-ditq8
last_updated: 2026-08-21
license: other
pipeline_tag: text-to-video-synthesis
tasks:
  - text-to-video-synthesis
base_model:
  - Lightricks/LTX-2.5
base_model_relation: finetune
parameters: 5.7B
tensor_type:
  - U32
  - F32
  - BF16
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 122
stars: 1
tags:
  - mlx
  - ltx-video
  - ltx-2.5
  - text-to-video
  - image-to-video
  - audio-video
  - quantized
---

# ltx-2.5-mlx-ditq8

> ltx-2.5-mlx-ditq8 - mlx-community 在 ModelScope 开源的模型。LTX-2.5 — int8 DiT (MLX)

mlx-community/ltx-2.5-mlx-ditq8 是 ModelScope 魔搭社区上的 5.7B 参数text-to-video-synthesis模型，采用 other 许可，基于 Lightricks/LTX-2.5 构建。

- **Repository**: mlx-community/ltx-2.5-mlx-ditq8
- **License**: other
- **Tasks**: text-to-video-synthesis
- **Parameters**: 5.7B
- **Base model**: Lightricks/LTX-2.5
- **Tags**: mlx, ltx-video, ltx-2.5, text-to-video, image-to-video, audio-video, quantized
- **Downloads**: 122
- **Stars**: 1
- **Last updated**: 2026-08-21

Source: https://www.modelscope.cn/models/mlx-community/ltx-2.5-mlx-ditq8

---

# LTX-2.5 — int8 DiT (MLX)

The **transformer only**, quantized to int8 group-64, for
[`mlx-community/ltx-2.5-mlx`](https://huggingface.co/mlx-community/ltx-2.5-mlx). Every other
component (Gemma-4 text encoder, connector, VAEs, vocoder, duration head, upscalers) comes from
that base repo — this repo deliberately carries **one file**, so there is nothing to keep in sync.

MLX conversion of [Lightricks LTX-2.5](https://huggingface.co/Lightricks/LTX-2.5), consumed by
[`ltx-2-mlx`](https://github.com/xocialize/ltx-2-mlx) (Python) and
[`ltx-2-mlx-swift`](https://github.com/xocialize/ltx-2-mlx-swift) (Swift).

## Why this exists

LTX-2.5's bf16 DiT is ~38 GB resident. On a 64 GB machine that left the tier it ships on at **96%
of its memory budget** — technically admissible, with 1.69 GB of headroom, at a geometry *below*
the profile's own frame cap. int8 takes the same tier to **57%**.

## Recipe

`int8`, **group 64**, `transformer_blocks` only — identical to the LTX-2.3 recipe, and read off the
shipped 2.3 q8 checkpoint rather than invented. `QUANT-RECIPE.json` in this repo records it
machine-readably.

⚠️ **There is no int4 sibling, deliberately.** int4 was built, gated and **rejected**: connector
output at valid token positions read 0.996728 against a 0.999879 bf16 floor — 27× the baseline
angular error. It is not published because it should not be used.

## Fidelity

Measured against the bf16 DiT on identical inputs at the shipping geometry (N=5632):
**cosine 0.99824** — and for calibration, LTX-2.3's q8 measured the *same way* reads 0.99238, so
2.5's int8 sits **4.3× closer to its bf16** than 2.3's does.

⚠️ A quant-vs-bf16 cosine is only readable against the same measurement on a known arm. Do not
compare this figure to numbers produced by a different harness.

## Use with weight streaming

Pairs with [`xocialize/ltx-2.5-granules`](https://huggingface.co/xocialize/ltx-2.5-granules) — a
per-block re-layout of these exact bytes that streams the DiT from SSD instead of loading it
resident. Output is **bit-identical** to the resident path (memcmp-gated, with a poisoned-slot
negative control). Streaming replaces this file rather than supplementing it: fetch one or the
other, not both.

## Licence

LTX-2 Community License — see `LICENSE.md` and the acceptable-use policy snapshot in this repo.
The weights are a derivative of Lightricks LTX-2.5 and carry its terms, including the revenue gate
and non-compete in Attachment A. Redistribution is permitted under §3 with those terms propagated.