fluently/FluentlyQwen3-Coder-4B-0909

Hugging Face
TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 4, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

FluentlyQwen3-Coder-4B-0909 is a 4 billion parameter causal language model developed by Project Fluently, based on the Qwen3 architecture. This model is specifically fine-tuned for generating high-quality code, leveraging SFT and GRPO training with step-by-step merging. It features a native context length of 32,768 tokens and can extend to 131,072 tokens with YaRN, making it suitable for extensive coding tasks and general improvements across various domains.

Loading preview...

FluentlyQwen3 Coder 4B Overview

FluentlyQwen3-Coder-4B-0909 is a 4 billion parameter causal language model from Project Fluently, built upon the Qwen3 architecture. It has been developed through Supervised Fine-Tuning (SFT) and Grouped Random Permutation Optimization (GRPO) training, incorporating a step-by-step merging process to enhance its capabilities. This model is particularly distinguished as a Coder Edition, specifically fine-tuned for generating high-quality code.

Key Capabilities & Features

  • Code Generation: Specially optimized for producing high-quality code.
  • Extended Context Length: Natively supports a 32,768-token context window, expandable to 131,072 tokens using YaRN.
  • Thinking Mode: Features an enable_thinking switch, allowing the model to engage in reasoning processes (similar to QwQ-32B) to improve response quality. This mode generates content within <think>...</think> blocks.
  • Non-Thinking Mode: A hard switch to disable the thinking behavior, aligning functionality with previous Qwen2.5-Instruct models for enhanced efficiency in specific scenarios.
  • General Improvements: Demonstrates improved performance across a range of tasks including basic communication, translation, mathematics, physics, biology, medicine, and agent functions.

When to Use This Model

This model is ideal for developers and applications requiring robust code generation capabilities. Its thinking mode can be beneficial for tasks demanding complex reasoning, while the non-thinking mode offers efficiency for straightforward generation. The extended context length makes it suitable for processing and generating longer code snippets or detailed technical documentation.

Finetunes
2 models