EAGLE3 For nex-agi/SGLANG-EAGLE3-Qwen3-32B-Nex-N1

About

Nex is a next-generation, full-stack agentic platform that brings foundation models, synthetic data pipelines, RL training, agent frameworks, and deployment tools together in one unified ecosystem.

SpecBundle is an open-source initiative, jointly driven by the community and industry, to democratize speculative decoding by providing high-performance speculative decoding draft weights for mainstream open-source models.

This checkpoint was trained by the Nex-AGI Team and released as the phase 1 of SpecBundle release. We regenerated the responses in the mlabonne/open-perfectblend and trained the model on 1.4M data samples.

Usage

You can use this checkpoint with the command below.

python3 -m sglang.launch_server \
    --model nex-agi/Qwen3-32B-Nex-N1 \
    --speculative-algorithm EAGLE3 \
    --speculative-draft-model-path nex-agi/SGLANG-EAGLE3-Qwen3-32B-Nex-N1 \
    --speculative-num-steps 3 \
    --speculative-eagle-topk 1 \
    --speculative-num-draft-tokens 4 \
    --tp 4

Performance

This checkpoint exhibits superior performance on various benchmarks.

Throughput	Acceptance Length

You can reproduce the performance with the command below:

# clone specforge
git clone https://github.com/sgl-project/SpecForge.git
cd SpecForge/benchmarks

# run benchmarks
python bench_eagle3.py \
        --model nex-agi/Qwen3-32B-Nex-N1 \
        --speculative-algorithm EAGLE3 \
        --speculative-draft-model-path nex-agi/SGLANG-EAGLE3-Qwen3-32B-Nex-N1 \
        --port 30002 \
        --config-list 8,0,0,0 8,3,1,4 8,5,1,6 8,5,3,6 8,7,1,8 8,7,4,8 \
        --benchmark-list gsm8k math500 mtbench humaneval livecodebench financeqa gpqa \
        --dtype bfloat16 \
        --tp 4 \
        --name Qwen3-32B-Nex-N1-spec-bundle