---
title: Qwen3-VL-235B-A22B-Instruct-speculator.eagle3
canonical_url: "https://www.modelscope.ai/models/RedHatAI/Qwen3-VL-235B-A22B-Instruct-speculator.eagle3"
md_url: "https://www.modelscope.ai/models/RedHatAI/Qwen3-VL-235B-A22B-Instruct-speculator.eagle3.md"
repository: RedHatAI/Qwen3-VL-235B-A22B-Instruct-speculator.eagle3
last_updated: 2026-08-29
license: apache-2.0
architectures:
  - Eagle3DraftModel
base_model:
  - Qwen/Qwen3-VL-235B-A22B-Instruct
base_model_relation: finetune
parameters: 1.2B
tensor_type:
  - BF16
  - BOOL
  - I64
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 79
stars: 0
tags:
  - eagle3
  - neuralmagic
  - speculators
  - qwen
---

# Qwen3-VL-235B-A22B-Instruct-speculator.eagle3

> Qwen3-VL-235B-A22B-Instruct-speculator.eagle3 - An open-source model by RedHatAI on ModelScope. Model Overview Verifier: Qwen/Qwen3-VL-235B-A22B-Instruct Speculative Decoding Algorithm: EAGLE-3 Model Architecture: Eagle3Speculator Release Date: 01/23/2026…

RedHatAI/Qwen3-VL-235B-A22B-Instruct-speculator.eagle3 is a 1.2B-parameter machine learning model on ModelScope. licensed under apache-2.0. derived from Qwen/Qwen3-VL-235B-A22B-Instruct.

- **Repository**: RedHatAI/Qwen3-VL-235B-A22B-Instruct-speculator.eagle3
- **License**: apache-2.0
- **Parameters**: 1.2B
- **Base model**: Qwen/Qwen3-VL-235B-A22B-Instruct
- **Tags**: eagle3, neuralmagic, speculators, qwen
- **Downloads**: 79
- **Stars**: 0
- **Last updated**: 2026-08-29

Source: https://www.modelscope.ai/models/RedHatAI/Qwen3-VL-235B-A22B-Instruct-speculator.eagle3

---

## Model Overview
- **Verifier:** Qwen/Qwen3-VL-235B-A22B-Instruct
- **Speculative Decoding Algorithm:** EAGLE-3
- **Model Architecture:** Eagle3Speculator
- **Release Date:** 01/23/2026
- **Version:** 1.0
- **Model Developers:** RedHat

This is a speculator model designed for use with [Qwen/Qwen3-VL-235B-A22B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Instruct), based on the [EAGLE-3](https://arxiv.org/abs/2503.01840) speculative decoding algorithm.
It was trained using the [speculators](https://github.com/neuralmagic/speculators) library on a combination of the [Magpie-Align/Magpie-Llama-3.1-Pro-300K-Filtered](https://huggingface.co/datasets/Magpie-Align/Magpie-Llama-3.1-Pro-300K-Filtered) dataset and the `train_sft` split of the [HuggingFaceH4/ultrachat_200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) dataset.
This model should be used with the [Qwen/Qwen3-VL-235B-A22B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Instruct) chat template, specifically through the `/chat/completions` endpoint.

## Use with vLLM

```bash
vllm serve Qwen/Qwen3-VL-235B-A22B-Instruct \
  -tp 8 \
  --speculative-config '{
    "model": "RedHatAI/Qwen3-VL-235B-A22B-Instruct-speculator.eagle3",
    "num_speculative_tokens": 3,
    "method": "eagle3"
  }'
```

## Evaluations

<h3>Use cases</h3>
<table>
  <thead>
    <tr>
      <th>Use Case</th>
      <th>Dataset</th>
      <th>Number of Samples</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>Coding</td>
      <td>HumanEval</td>
      <td>168</td>
    </tr>
    <tr>
      <td>Math Reasoning</td>
      <td>gsm8k</td>
      <td>80</td>
    </tr>
    <tr>
      <td>Text Summarization</td>
      <td>CNN/Daily Mail</td>
      <td>80</td>
    </tr>
  </tbody>
</table>
<h3>Acceptance lengths</h3>
<table>
  <thead>
    <tr>
      <th>Use Case</th>
      <th>k=1</th>
      <th>k=2</th>
      <th>k=3</th>
      <th>k=4</th>
      <th>k=5</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>Coding</td>
      <td>1.85</td>
      <td>2.49</td>
      <td>2.88</td>
      <td>3.13</td>
      <td>3.24</td>
    </tr>
    <tr>
      <td>Math Reasoning</td>
      <td>1.85</td>
      <td>2.54</td>
      <td>2.96</td>
      <td>3.25</td>
      <td>3.24</td>
    </tr>
    <tr>
      <td>Text Summarization</td>
      <td>1.63</td>
      <td>1.94</td>
      <td>2.08</td>
      <td>2.14</td>
      <td>2.16</td>
    </tr>
  </tbody>
</table>
<details> <summary>Details</summary>
<strong>Configuration</strong>

- repetitions: 1
- time per experiment: 10min
- hardware: 8xA100
- vLLM version: 0.11.2
- GuideLLM version: 0.3.0

<strong>Command</strong>
```bash
GUIDELLM__PREFERRED_ROUTE="chat_completions" \
guidellm benchmark \
  --target "http://localhost:8000/v1" \
  --data "RedHatAI/speculator_benchmarks" \
  --data-args '{"data_files": "HumanEval.jsonl"}' \
  --rate-type sweep \
  --max-seconds 600 \
  --output-path "Qwen235B-HumanEval.json" \
```
GuideLLM interface changed, so for compatibility with the latest version (v0.6.0), please use the following command:
```bash
GUIDELLM__PREFERRED_ROUTE="chat_completions" \
guidellm benchmark \
  --target "http://localhost:8000/v1" \
  --data "RedHatAI/speculator_benchmarks" \
  --data-args '{"data_files": "HumanEval.jsonl"}' \
  --profile sweep \
  --max-seconds 1800 \
  --output-path "my_output.json" \
  --backend-args '{"extras": {"body": {"temperature":0.6, "top_p":0.95, "top_k":20}}}'
```
</details>