yothinS/gemma4-thai-marketing

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:5.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 24, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

yothinS/gemma4-thai-marketing is a 5.1 billion parameter Gemma 4 E2b architecture model, fine-tuned by yothinS using QLoRA and Unsloth Core. It specializes in analyzing performance marketing data, particularly for Meta Ads and TikTok Ads, and providing IF-THEN actionable insights in Thai. The model is optimized for rapid inference and processing of Thai language responses, focusing on business data analysis.

Loading preview...

Overview

yothinS/gemma4-thai-marketing is a 5.1 billion parameter Gemma 4 E2b model, fine-tuned by yothinS using QLoRA via Unsloth Core. This model is specifically designed for Performance Marketing analysis, focusing on identifying bottlenecks in advertising campaigns (Meta Ads, TikTok Ads) and providing structured, actionable insights in Thai. It leverages a specialized dataset of online marketing statistics and concepts, cleaned and formatted for chat-based interactions.

Key Capabilities

  • Expert in Performance Marketing: Provides in-depth analysis of ad campaigns, including root cause identification and actionable steps.
  • Optimized for Thai: Developed with Unsloth Core for fast inference and processing of Thai language responses.
  • Flexible Usage: Available as a merged 16-bit model for direct use with Transformers or as a LoRA Adapter for memory-efficient deployment with Unsloth.
  • Domain-Specific Insights: Trained on metrics like CPM, CTR, CPC, CVR, and ROAS to offer relevant business recommendations.

Good For

  • Analyzing performance marketing data and identifying issues in ad campaigns.
  • Generating structured, IF-THEN actionable insights for online advertising.
  • Users requiring rapid, specialized Thai language processing for marketing analytics.

Limitations

  • Domain Specificity: The model's responses are heavily biased towards performance marketing, potentially leading to a formal or IF-THEN structured style for general queries.
  • Verification Required: Recommendations and statistical figures should be cross-referenced with actual ad dashboard data (e.g., Meta/TikTok Event Manager) before making business decisions.

Finetunes
2 models