ermiaazarkhalili/Qwen3-4B-Function-Calling-xLAM-Unsloth

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Apr 25, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The ermiaazarkhalili/Qwen3-4B-Function-Calling-xLAM-Unsloth model is a 4 billion parameter Qwen3 variant, fine-tuned by ermiaazarkhalili for function calling. It leverages Unsloth for 2x faster training and 60% less VRAM, trained on the Salesforce/xlam-function-calling-60k dataset. This model excels at interpreting natural language queries and generating structured function calls, making it suitable for tool-use applications. It has a context length of 2,048 tokens and is available in GGUF formats for efficient deployment.

Loading preview...

Model Overview

This model, ermiaazarkhalili/Qwen3-4B-Function-Calling-xLAM-Unsloth, is a specialized 4-billion parameter variant of the Qwen3 architecture, developed by ermiaazarkhalili. It is specifically fine-tuned for function calling capabilities, enabling it to translate natural language requests into structured tool invocations.

Key Capabilities & Features

  • Optimized Function Calling: Fine-tuned on the Salesforce/xlam-function-calling-60k dataset, comprising 60,000 examples of queries, tool definitions, and structured answers.
  • Efficient Training: Utilizes Unsloth for training, resulting in 2x faster training and 60% less VRAM consumption compared to standard methods.
  • QLoRA Fine-tuning: Employs 4-bit QLoRA with a rank of 16, targeting key attention and feed-forward modules for efficient adaptation.
  • Compact Size: With 4 billion parameters and a 2,048-token context window, it offers a balance of performance and resource efficiency.
  • Deployment Flexibility: Available in various GGUF formats (Q4_K_M, Q5_K_M, Q8_0) for CPU and edge device inference, compatible with tools like Ollama and llama.cpp.

Ideal Use Cases

  • Tool-Use Agents: Developing AI agents that can interact with external APIs and tools based on user prompts.
  • Structured Output Generation: Applications requiring the model to output data in a predefined, structured format (e.g., JSON for function calls).
  • Resource-Constrained Environments: Its Unsloth optimization and GGUF availability make it suitable for deployment on devices with limited GPU or CPU resources.

Limitations

  • Primarily trained on English data.
  • Knowledge cutoff is limited to the base model's training data.
  • May exhibit hallucinations and is not extensively safety-tuned, requiring appropriate guardrails in production.

Finetunes
2 models