amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-context

Model Overview

P-EAGLE is a parallel-drafting speculative decoding model that generates K draft tokens in a single forward pass. It transforms EAGLE—the state-of-the-art speculative decoding method—from autoregressive to parallel draft generation.

For use cases of less than 10k context length - please consider using Qwen3-Coder-30B-A3B-Instruct-P-EAGLE.

Model Details

The model architecture is illustrated in the following figure. Specifically, we trained a 4-layer P-EAGLE for Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 as the target model, with number of parallel-token prediction as 18.

P-EAGLE follows the vanila EAGLE 3 using three layers of hidden states from the target model.

Model Description

• Developed by: AWS
• Model type: EAGLE
• Language(s) (NLP): English
• License: Apache License 2.0
• Target model: Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8

Model Sources

• Paper: P-EAGLE: Parallel-Drafting EAGLE with Scalable Training

Training Data

• nvidia/OpenCodeInstruct 200K
• Ultrachat_200k

Similar to nvidia/gpt-oss-120b-Eagle3-long-context: only prompts from the datasets were used for data synthesis (the original responses from GPT were not used for data synthesis) which is then used to train the P-Eagle.

Usage

To serve the checkpoint in vLLM

vllm serve \
--model Qwen/Qwen3-Coder-30B-A3B-Instruct \
--tensor-parallel-size 1 \
--speculative-config '{"method": "eagle3", "model": "amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-context", "num_speculative_tokens": 10, "parallel_drafting": true}' \
--no-enable-prefix-caching \
--async-scheduling \
--max-model-len 131072 \
--max-num-batched-tokens 131072

Evaluation

From vllm-bench, with speculation length of 10 and max-new-token of 2048, we see the following acceptance length for Qwen3-Coder-30B-A3B-Instruct-P-EAGLE and its long-context extension (the current model).

Benchmark	Qwen3-Coder-30B-A3B-Instruct-P-EAGLE	long-context extension
aa-lcr (100k)	1.20	2.19
humaneval	5.30	5.01
mtbench	3.25	2.65

vLLM bench command is shown as below.

vllm bench serve \
      --backend openai-chat \
      --endpoint /v1/chat/completions \
      --model Qwen/Qwen3-Coder-30B-A3B-Instruct \
      --dataset-name custom \
      --dataset-path /home/ubuntu/eval_datasets/humaneval_custom.jsonl \
      --custom-output-len 256 \
      --num-prompts 80 \
      --max-concurrency 1 \
      --temperature 0 \
      --request-rate inf \
      --save-result --save-detailed

Ciatation

@article{hui2026p,
  title={P-EAGLE: Parallel-Drafting EAGLE with Scalable Training},
  author={Hui, Mude and Huang, Xin and Salas, Jaime Campos and Sun, Yue and Pemberton, Nathan and Song, Xiang and Khetan, Ashish and Karypis, George},
  journal={arXiv preprint arXiv:2602.01469},
  year={2026}
}