NotoriousH2/Qwen3-0.6B-JSON-SFT-GRPO

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.8BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 19, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

NotoriousH2/Qwen3-0.6B-JSON-SFT-GRPO is a 0.8 billion parameter Qwen3 model developed by NotoriousH2, fine-tuned for JSON instruction following. This model was trained using Unsloth and Huggingface's TRL library, enabling faster fine-tuning. It is designed for tasks requiring structured JSON output, leveraging its base Qwen3 architecture and specialized instruction tuning.

Loading preview...

Overview

NotoriousH2/Qwen3-0.6B-JSON-SFT-GRPO is a specialized 0.8 billion parameter Qwen3 model, developed by NotoriousH2, that has been fine-tuned for generating structured JSON outputs. This model builds upon the NotoriousH2/Qwen3-0.6B-JSON-SFT base and was trained with enhanced efficiency using the Unsloth library in conjunction with Huggingface's TRL library, resulting in a 2x speed improvement during fine-tuning.

Key Capabilities

  • JSON Instruction Following: Optimized to understand and respond to prompts requiring JSON formatted output.
  • Efficient Training: Benefits from Unsloth's optimizations for faster fine-tuning, making it suitable for rapid iteration and deployment.
  • Compact Size: At 0.8 billion parameters, it offers a balance between performance and computational efficiency.

Good For

  • Applications requiring reliable, structured JSON responses from a language model.
  • Integration into systems where a smaller, efficient model is preferred for JSON generation tasks.
  • Developers looking for a Qwen3 variant specifically tailored for JSON output with efficient training origins.

Finetunes
1 models