zenlm/zen3-reranker-medium

Zen3 Reranker Medium

Medium-sized Zen3 reranker for higher accuracy re-scoring in RAG pipelines.

Overview

Built on Zen MoDE (Mixture of Distilled Experts) architecture with 1.5B parameters and 8K context window.

Developed by Hanzo AI and the Zoo Labs Foundation.

Quick Start

from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

model_id = "zenlm/zen3-reranker-medium"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSequenceClassification.from_pretrained(model_id, torch_dtype="auto")

query = "What is machine learning?"
passages = [
    "Machine learning is a subset of artificial intelligence...",
    "The weather today is sunny and warm.",
    "Neural networks learn from labeled training data.",
]

pairs = [[query, p] for p in passages]
features = tokenizer(pairs, padding=True, truncation=True, max_length=512, return_tensors="pt")

with torch.no_grad():
    scores = model(**features).logits.squeeze(-1)

ranked = sorted(zip(scores.tolist(), passages), reverse=True)
for score, passage in ranked:
    print(f"Score: {score:.4f} | {passage[:80]}")

Model Details

Attribute	Value
Parameters	1.5B
Architecture	Zen MoDE
Context	8K tokens
License	Apache 2.0

License

Apache 2.0