---
title: DeepSeek-V4-Flash-MTP-bf16
canonical_url: "https://www.modelscope.cn/models/mlx-community/DeepSeek-V4-Flash-MTP-bf16"
md_url: "https://www.modelscope.cn/models/mlx-community/DeepSeek-V4-Flash-MTP-bf16.md"
repository: mlx-community/DeepSeek-V4-Flash-MTP-bf16
last_updated: 2026-06-02
license: mit
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - deepseek_v4_mtp
base_model:
  - deepseek-ai/DeepSeek-V4-Flash
base_model_relation: quantized
parameters: 1.1B
tensor_type:
  - F32
  - U32
  - U8
  - BF16
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
downloads: 83
stars: 0
tags:
  - mlx
  - mlx-vlm
  - deepseek_v4_mtp
  - deepseek-v4
  - deepseek-v4-flash
  - mtp
  - speculative-decoding
  - draft-model
  - bf16
  - 8-bit
---

# DeepSeek-V4-Flash-MTP-bf16

> DeepSeek-V4-Flash-MTP-bf16 - mlx-community 在 ModelScope 开源的模型。DeepSeek-V4-Flash-MTP-bf16

mlx-community/DeepSeek-V4-Flash-MTP-bf16 是 ModelScope 魔搭社区上的 1.1B 参数text-generation模型，采用 mit 许可，基于 deepseek-ai/DeepSeek-V4-Flash 构建。

- **Repository**: mlx-community/DeepSeek-V4-Flash-MTP-bf16
- **License**: mit
- **Tasks**: text-generation
- **Parameters**: 1.1B
- **Base model**: deepseek-ai/DeepSeek-V4-Flash
- **Tags**: mlx, mlx-vlm, deepseek_v4_mtp, deepseek-v4, deepseek-v4-flash, mtp, speculative-decoding, draft-model, bf16, 8-bit
- **Downloads**: 83
- **Stars**: 0
- **Last updated**: 2026-06-02

Source: https://www.modelscope.cn/models/mlx-community/DeepSeek-V4-Flash-MTP-bf16

---

# DeepSeek-V4-Flash-MTP-bf16

This repository contains Multi-Token Prediction (MTP) drafter weights split from `deepseek-ai/DeepSeek-V4-Flash` for use with `mlx-vlm` speculative decoding.

This is not a standalone chat or text-generation model. Load it as the draft model alongside a compatible DeepSeek V4 Flash target checkpoint.

## Use with mlx-vlm

```bash
uv run mlx_vlm.generate \
  --model mlx-community/DeepSeek-V4-Flash-4bit \
  --draft-model mlx-community/DeepSeek-V4-Flash-MTP-bf16 \
  --prompt "Hi, how are you?" \
  --max-tokens 256 \
  --enable-thinking
```

For local weights:

```bash
uv run mlx_vlm.generate \
  --model /path/to/target-model \
  --draft-model /path/to/DeepSeek-V4-Flash-MTP \
  --prompt "Hi, how are you?" \
  --max-tokens 256 \
  --enable-thinking
```

## Model Details

- Model type: `deepseek_v4_mtp`
- MTP block size: `2`
- Target architecture: DeepSeek V4 Flash
- Precision: bf16 / mixed 8-bit + 4-bit as configured
- Runtime: MLX / `mlx-vlm`
- Format: Safetensors with MLX-compatible config and tokenizer files

The stored tensors include bfloat16 parameters and MLX quantized tensors as described in `config.json`.

## Intended Use

Use this repo only as a speculative decoding drafter for compatible DeepSeek V4 Flash checkpoints. The target model verifies drafted tokens, while this MTP model proposes candidate tokens per decoding step.

## Limitations

This checkpoint requires runtime support for Qwen/DeepSeek MTP draft models in `mlx-vlm`. Standard standalone generation through generic Transformers APIs is not expected to work with this repository by itself.

Please refer to the upstream `deepseek-ai/DeepSeek-V4-Flash` model card and license terms for model usage constraints.