Qwen3-4B-Esper3 Open Source Coding Model - Essential for Local and Server Deployment, Development and Operations Support

Qwen3 4B Esper3

Developed by ValiantLabs

Esper 3 is an expert model for coding, architecture, and DevOps inference built upon Qwen3, suitable for local and server deployment.

Large Language Model

Transformers

Supports Multiple LanguagesOpen Source License:Apache-2.0 #DevOps Inference #Code Generation Expert #Cloud Architecture Design

Downloads 4,951

Release Time : 5/6/2025

Model Overview

Esper 3 is a specialized inference expert model focused on code generation, architecture design, and DevOps tasks, fine-tuned from Qwen3-4B, supporting rapid inference and problem-solving.

Model Features

DevOps and Architecture Inference

Fine-tuned with DevOps and architecture inference data, excels at solving problems in related fields.

Code Reasoning Capability

Enhanced with code reasoning data, capable of handling code generation tasks across multiple programming languages.

General and Creative Reasoning

Improved general reasoning abilities enhance problem-solving and general chat performance.

Efficient Inference

Small model size allows for operation on local desktops and mobile devices, supporting ultra-fast server inference.

Model Capabilities

Text Generation

Code Generation

Architecture Design

DevOps Automation

Problem Solving

Creative Reasoning

Use Cases

Cloud Computing Infrastructure

Terraform Configuration Generation

Automatically generates Terraform configuration code for AWS infrastructure.

Dynamically determines AMI ID and configures EC2 instances.

Software Development

Code Assistance

Helps developers write code in languages like Python and JavaScript.

Improves development efficiency and reduces repetitive work.

System Architecture

Architecture Design Consultation

Provides advice on cloud computing architecture design.

Helps optimize cloud resource usage and system architecture.

🚀 Esper 3: A Coding and Reasoning Specialist

Esper 3 is a powerful model specializing in coding, architecture, and DevOps reasoning, built upon the Qwen 3 framework. It offers high - performance reasoning capabilities, suitable for various programming and problem - solving tasks.

🚀 Quick Start

Prerequisites

Ensure you have the transformers library installed.

Example Inference Script

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "ValiantLabs/Qwen3-4B-Esper3"

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

# prepare the model input
prompt = "Write a Terraform configuration that uses the `aws_ami` data source to find the latest Amazon Linux 2 AMI. Then, provision an EC2 instance using this dynamically determined AMI ID."
messages = [
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=True # Switches between thinking and non-thinking modes. Default is True.
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

# conduct text completion
generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=32768
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist() 

# parsing thinking content
try:
    # rindex finding 151668 (</think>)
    index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
    index = 0

thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")

print("thinking content:", thinking_content)
print("content:", content)

✨ Features

Finetuned on Specialized Data: Esper 3 is finetuned on DevOps and architecture reasoning and code reasoning data generated with Deepseek R1, enhancing its performance in relevant fields.
Improved Reasoning Ability: It has improved general and creative reasoning to supplement problem - solving and general chat performance.
Small Model Size: The small model sizes allow it to run on local desktops and mobile devices, and also support super - fast server inference.