---
title: Minimax-M3.1-Flash-Dflash2
canonical_url: "https://www.modelscope.cn/models/RadixArk/Minimax-M3.1-Flash-Dflash2"
md_url: "https://www.modelscope.cn/models/RadixArk/Minimax-M3.1-Flash-Dflash2.md"
repository: RadixArk/Minimax-M3.1-Flash-Dflash2
last_updated: 2026-10-23
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - qwen3
architectures:
  - MiniMaxDFlash2DraftModel
base_model:
  - MiniMaxAI/MiniMax-M3.1-Flash
base_model_relation: finetune
parameters: 2.1B
tensor_type:
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 6
stars: 0
tags:
  - speculative-decoding
  - dflash2
  - minimax
---

# Minimax-M3.1-Flash-Dflash2

> Minimax-M3.1-Flash-Dflash2 - RadixArk 在 ModelScope 开源的模型。Minimax-M3.1-Flash-Dflash2

RadixArk/Minimax-M3.1-Flash-Dflash2 是 ModelScope 魔搭社区上的 2.1B 参数text-generation模型，基于 MiniMaxAI/MiniMax-M3.1-Flash 构建。

- **Repository**: RadixArk/Minimax-M3.1-Flash-Dflash2
- **Tasks**: text-generation
- **Parameters**: 2.1B
- **Base model**: MiniMaxAI/MiniMax-M3.1-Flash
- **Tags**: speculative-decoding, dflash2, minimax
- **Downloads**: 6
- **Stars**: 0
- **Last updated**: 2026-10-23

Source: https://www.modelscope.cn/models/RadixArk/Minimax-M3.1-Flash-Dflash2

---

# Minimax-M3.1-Flash-Dflash2

## Overview

A DFlash2 draft for [MiniMax M3.1 Flash](https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash) that proposes tokens in parallel for verification by the target model in [SGLang](https://github.com/sgl-project/sglang).

This repository provides the BF16 checkpoint and its configuration. Serving also requires the target model, which provides the embedding, output head, tokenizer, and native chat template.

## Model Specifications

- Base model: `MiniMaxAI/MiniMax-M3.1-Flash`
- Architecture: `MiniMaxDFlash2DraftModel`, 5 GQA layers
- Parameters: 2,054,677,760 across 81 BF16 tensors
- Format: One BF16 Safetensors file
- Hidden size: 6,144; 64 query heads and 4 KV heads
- Attention: Sliding window of 4,096 tokens
- Verification width: 8 tokens, comprising an anchor and 7 proposals
- Selector: Rank 256, top-k 16
- Auxiliary target layers: `[3, 17, 31, 45, 59]`
- Configured maximum positions: 1,048,576

The evaluations below used a 16,384-token serving context. They do not establish acceptance quality at the full configured context length.

## Evaluation Results

`acc_len` is the unweighted mean of each request's completion-token count divided by its number of verification steps. It includes responses that reached the output cap.

| Dataset | Samples | acc_len |
| --- | ---: | ---: |
| GSM8K | 1,319 | 5.163813 |
| AIME26 | 30 | 4.309412 |
| MBPP | 257 | 4.383587 |
| MT-Bench | 160 turns (80 conversations) | 3.580651 |

Measured with `temperature=1.0`, `top_p=0.95`, `top_k=-1`, and native adaptive thinking. The evaluation used TP4 on GB300, a native NVFP4 target with TRTLLM MHA and FP8 KV cache, and a BF16 draft with FA4 and BF16 KV cache. DFLASH verified 8 tokens per step.

GSM8K, AIME26, and MBPP used concurrency 16 and an 8,192-token output cap; MT-Bench used concurrency 2 and a 2,048-token cap. The cap was reached by 3, 8, 3, and 54 responses, respectively. MT-Bench's second turn used the model's generated first-turn history. Acceptance length measures decoding behavior, rather than answer quality or throughput.

## Serving with SGLang

The following [cookbook](https://docs.sglang.io/cookbook/autoregressive/MiniMax/MiniMax-M3.1) and example show how to serve MiniMax M3.1 Flash with this DFlash2 draft using SGLang:

```bash
SGLANG_RAGGED_VERIFY_MODE=static \
FMHA_SM100_ALLOW_JIT=1 \
SGLANG_OPT_USE_BF16_ROUTER_GEMM=0 \
SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK=32768 \
python -m sglang.launch_server \
  --model-path MiniMaxAI/MiniMax-M3.1-Flash \
  --trust-remote-code \
  --reasoning-parser auto \
  --tool-call-parser auto \
  --tp-size 4 \
  --dp-size 1 \
  --ep-size 1 \
  --context-length 16384 \
  --chunked-prefill-size 2048 \
  --max-prefill-tokens 65536 \
  --max-running-requests 16 \
  --cuda-graph-max-bs-decode 16 \
  --cuda-graph-backend-prefill disabled \
  --mem-fraction-static 0.8 \
  --max-total-tokens 131072 \
  --moe-runner-backend flashinfer_trtllm_routed \
  --moe-a2a-backend none \
  --speculative-algorithm DFLASH \
  --speculative-draft-model-path RadixArk/Minimax-M3.1-Flash-Dflash2 \
  --speculative-draft-attention-backend fa4 \
  --speculative-draft-kv-cache-dtype bf16 \
  --speculative-draft-model-quantization unquant \
  --speculative-num-draft-tokens 8 \
  --random-seed 42 \
  --host 127.0.0.1 \
  --port 30000
```

SGLang selects the target's native NVFP4 loading, TRTLLM MHA attention, and FP8 KV cache automatically. Use the target's native chat template and adaptive thinking to match the evaluation settings.