RedHatAI/Llama4-Maverick-17B-128E-Instruct-speculator.eagle3

Llama-4-Maverick-17B-128E-Instruct-speculators.eagle3

Model Overview

• Verifier: meta-llama/Llama-4-Maverick-17B-128E-Instruct
• 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 meta-llama/Llama-4-Maverick-17B-128E-Instruct, based on the EAGLE-3 speculative decoding algorithm. It was converted into the speculators format from the model nvidia/Llama-4-Maverick-17B-128E-Eagle3. This model should be used with the meta-llama/Llama-4-Maverick-17B-128E-Instruct chat template, specifically through the /chat/completions endpoint.

Use with vLLM

vllm serve meta-llama/Llama-4-Maverick-17B-128E-Instruct \
  -tp 8 \
  --speculative-config '{
    "model": "RedHatAI/Llama-4-Maverick-17B-128E-Instruct-speculators.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	1.83	2.45	2.94	3.26	3.47	3.57	3.62
Math Reasoning	1.86	2.56	3.08	3.53	3.73	3.91	4.02
Text Summarization	1.69	2.12	2.37	2.52	2.60	2.63	2.63

Performance benchmarking (2xA100)

Details

Configuration

• temperature: 0.6
• top_p: 0.9
• repetitions: 3
• time per experiment: 3min
• hardware: 8xB200
• 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 180 \
  --output-path "Llama-4-Maverick-HumanEval.json" \
  --backend-args '{"extra_body": {"chat_completions": {"temperature":0.6, "top_p":0.9}}}'

Citation

If you use this model, please cite both the original NVIDIA model and the Speculators library:

@misc{nvidia2025llama4maverick,
  title={Llama 4 Maverick 17B Eagle3},
  author={NVIDIA Corporation},
  year={2025},
  publisher={Hugging Face}
}

@misc{speculators2024,
  title={Speculators: A Unified Library for Speculative Decoding},
  author={Neural Magic},
  year={2024},
  url={https://github.com/neuralmagic/speculators}
}

Acknowledgments

• Original model by NVIDIA Corporation
• Conversion and formatting for Speculators/vLLM compatibility
• Based on Eagle3 architecture with Llama3 draft head targeting Llama4 verifier