Qwen3-32B-speculator.eagle3

Model Overview

• Verifier: Qwen/Qwen3-32B
• Speculative Decoding Algorithm: EAGLE-3
• Model Architecture: Eagle3Speculator
• Release Date: 09/17/2025
• Version: 1.0
• Model Developers: RedHat

This is a speculator model designed for use with Qwen/Qwen3-32B, based on the EAGLE-3 speculative decoding algorithm. It was trained using the speculators library on a combination of the Aeala/ShareGPT_Vicuna_unfiltered and the train_sft split of HuggingFaceH4/ultrachat_200k datasets. This model should be used with the Qwen/Qwen3-32B chat template, specifically through the /chat/completions endpoint.

Use with vLLM

vllm serve Qwen/Qwen3-32B \
  -tp 2 \
  --speculative-config '{
    "model": "RedHatAI/Qwen3-32B-speculator.eagle3",
    "num_speculative_tokens": 3,
    "method": "eagle3"
  }'

Evaluations

Use cases

Use Case	Dataset	Number of Samples
Coding	HumanEval	168
Math Reasoning	gsm8k	80
Text Summarization	CNN/Daily Mail	80

Acceptance lengths

Use Case	k=1	k=2	k=3	k=4	k=5	k=6	k=7
Coding	1.67	2.06	2.29	2.39	2.47	2.50	2.53
Math Reasoning	1.73	2.21	2.49	2.69	2.80	2.83	3.08
Text Summarization	1.62	1.95	2.15	2.23	2.27	2.32	2.33

Performance benchmarking (2xA100)

Details

Configuration

• temperature: 0.6
• top_p: 0.95
• top_k: 20
• repetitions: 3
• time per experiment: 10min
• hardware: 2xA100
• vLLM version: 0.11.0
• GuideLLM version: 0.3.0

Command

GUIDELLM__PREFERRED_ROUTE="chat_completions" \
guidellm benchmark \
  --target "http://localhost:8000/v1" \
  --data "RedHatAI/SpeculativeDecoding" \
  --rate-type sweep \
  --max-seconds 600 \
  --output-path "Qwen3-32B-HumanEval.json" \
  --backend-args '{"extra_body": {"chat_completions": {"temperature":0.6, "top_p":0.95, "top_k":20}}}'

</details>