Neura-Tech-AI/Neuron-4B-Instruct

TEXT GENERATIONConcurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 25, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Neuron-4B-Instruct is a 4 billion parameter instruction-tuned language model developed by Neura Tech AI, built upon the Qwen/Qwen3-4B-Instruct-2507 base model. It features a 262,144-token context length and excels in reasoning, coding assistance, and multilingual understanding across numerous languages. This model is designed for a wide range of applications including AI assistants, chatbots, and general NLP tasks.

Loading preview...

Neuron-4B-Instruct Overview

Neuron-4B-Instruct is an instruction-tuned large language model developed by Neura Tech AI, based on the Qwen/Qwen3-4B-Instruct-2507 architecture. This model aims to provide a robust open-source AI assistant with strong capabilities in natural conversation, coding, reasoning, and multilingual processing.

Key Capabilities & Features

  • Architecture: Transformer Decoder with approximately 4 billion parameters.
  • Context Length: Inherits a substantial 262,144-token context window from its base model.
  • Multilingual Support: Supports a broad array of languages including English, Hindi, Chinese, Japanese, Korean, French, German, Spanish, and many others, leveraging the Qwen3 base model's capabilities.
  • Diverse Functionality: Offers instruction following, chat assistance, coding support, mathematical and logical reasoning, creative writing, and general knowledge processing.
  • Tool Calling: Includes support for tool calling, enhancing its utility in agentic workflows.

Performance Highlights

Neuron-4B-Instruct demonstrates competitive performance across various benchmarks, often outperforming its base model and other comparably sized models in specific areas:

  • Knowledge: Achieves 69.6 on MMLU-Pro and 84.2 on MMLU-Redux, surpassing Qwen3-4B.
  • Reasoning: Shows significant strength with 47.4 on AIME25 and 80.2 on ZebraLogic.
  • Coding: Scores 35.1 on LiveCodeBench v6 and 76.8 on MultiPL-E.
  • Alignment: Excels in creative writing with 83.5 on Creative Writing v3 and 83.4 on WritingBench.
  • Agentic Tasks: Leads in BFCL-v3 with 61.9 and TAU1-Retail with 48.7.

Intended Use Cases

This model is well-suited for a variety of applications:

  • AI Assistants and Chatbots
  • Coding and Software Development Support
  • Educational Tools and Research
  • Content Generation and Translation
  • Complex Reasoning Tasks
  • Applications requiring long-context understanding and tool integration.

Neuron-4B-Instruct is released under the Apache-2.0 License, consistent with its Qwen3 base model.

Finetunes
1 models