Zen Eco 4B Agent

Tool-calling and function execution model. Part of the Zen Eco family.

Overview

Zen Eco 4B Agent is the agent variant of Zen Eco 4B with built-in tool-calling and function execution capabilities. Trained for structured tool use, JSON output, and multi-step agentic workflows.

Property	Value
Parameters	4B
Context	32K
License	Apache 2.0

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("zenlm/zen-eco-agent")
tokenizer = AutoTokenizer.from_pretrained("zenlm/zen-eco-agent")

tools = [
    {"type": "function", "function": {"name": "get_weather", "parameters": {"type": "object", "properties": {"location": {"type": "string"}}}}}
]

messages = [{"role": "user", "content": "What's the weather in Tokyo?"}]
inputs = tokenizer.apply_chat_template(messages, tools=tools, return_tensors="pt")
output = model.generate(inputs, max_new_tokens=256)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Related

• zen-eco — Base 4B model
• zen-eco-instruct — Instruction-tuned variant
• zen-eco-thinking — Chain-of-thought variant
• Zen LM — Full model family

Apache 2.0 · Zen LM · Hanzo AI