SnowGemma/Qwen2.5-1.5B-X-Writer-v1

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Oct 15, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

SnowGemma/Qwen2.5-1.5B-X-Writer-v1 is a 1.5 billion parameter language model based on Qwen/Qwen2.5-1.5B-Instruct, fine-tuned by SnowGemma. It specializes in generating Chinese X (Twitter) short posts and replies in a style closely matching the author's personal tone. This model is specifically designed for generating short-form social media content based on provided themes and materials, making it suitable for personalized content creation rather than general writing tasks. It is provided in MLX format with bfloat16 weights, optimized for direct loading on Apple Silicon Macs.

Loading preview...

Overview

SnowGemma/Qwen2.5-1.5B-X-Writer-v1 is a specialized 1.5 billion parameter language model, fine-tuned from Qwen/Qwen2.5-1.5B-Instruct. Its primary function is to generate Chinese X (formerly Twitter) short posts and replies that mimic the personal writing style and tone of its creator. This model is distributed in MLX format with bfloat16 weights, enabling direct use on Apple Silicon Macs.

Key Capabilities

  • Personalized Short-Form Content Generation: Excels at producing Chinese X short posts and replies based on user-provided themes and specific content requirements.
  • Style Emulation: Designed to replicate a specific author's personal tone and common topics, making it ideal for niche content creation.
  • Optimized for Apple Silicon: Provided in MLX format for efficient loading and inference on Apple Silicon Macs, with a generation speed of approximately 65 tokens/s and peak memory usage of 3.3GB on an M4 Pro.

Good for

  • Generating personalized social media updates: Users who need to create short, stylistically consistent Chinese X posts or replies.
  • Experimenting with LoRA fine-tuning: Serves as a practical example for learning and applying LoRA fine-tuning techniques on small models.
  • Content creation requiring a specific voice: Ideal for scenarios where the output needs to match a particular individual's online persona for short-form text.

Limitations

Due to its small size and limited training data (297 samples of short posts and replies), the model's long-form writing capabilities are similar to its base model. It may still hallucinate facts or numbers not present in the provided material, requiring human review. Its effectiveness is best when generating content in the author's specific style and topics; performance may vary for other domains or styles.

Finetunes
1 models