zenlm/zen-next-80b-instruct

Zen-Next (80B)

Part of the Zen AI Model Family

Model Description

Parameters: 80B
Base Model: Qwen/Qwen2.5-72B-Instruct
Specialization: Complex reasoning & extended context
Training: Flagship training with constitutional AI
Context: 32K-128K tokens
Thinking: Up to 1,000,000 tokens

Files in This Repository

This repository contains ALL formats and quantizations:

🔷 SafeTensors (Original)

• model.safetensors - Full precision weights
• config.json - Model configuration
• tokenizer.json - Fast tokenizer

🟢 GGUF Quantized

• zen-next-80b-instruct-Q4_K_M.gguf - 4-bit (recommended)
• zen-next-80b-instruct-Q5_K_M.gguf - 5-bit (balanced)
• zen-next-80b-instruct-Q8_0.gguf - 8-bit (high quality)

🍎 MLX (Apple Silicon)

• mlx-4bit/ - 4-bit quantized for M-series
• mlx-8bit/ - 8-bit quantized for M-series

Performance

Benchmark	Score	Rank
MMLU	75.6%	Top 10%
GSM8K	82.1%	Top 15%
HumanEval	61.7%	Top 20%

Quick Start

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("zenlm/zen-next-80b-instruct")
tokenizer = AutoTokenizer.from_pretrained("zenlm/zen-next-80b-instruct")

# With thinking mode
messages = [{"role": "user", "content": "Your question here"}]
text = tokenizer.apply_chat_template(messages, enable_thinking=True)

GGUF with llama.cpp

./main -m zen-next-80b-instruct-Q4_K_M.gguf -p "Your prompt" -n 512

MLX for Apple Silicon

from mlx_lm import load, generate
model, tokenizer = load("zenlm/zen-next-80b-instruct")
response = generate(model, tokenizer, "Your prompt", max_tokens=200)

Unique Training Background

Flagship training with constitutional AI

This model was specifically optimized for complex reasoning & extended context with careful attention to:

• Inference efficiency
• Memory footprint
• Quality preservation
• Thinking capabilities

---

Part of the Zen Family • Collection • GitHub