Kelnux/Qwen3-0.6B-seo-bilingual

Qwen3-0.6B SEO Bilingual (PL+EN)

A bilingual (Polish + English) SEO expert model fine-tuned from Qwen/Qwen3-0.6B using LoRA.

Model Details

• Base Model: Qwen/Qwen3-0.6B
• Fine-tuning Method: LoRA (r=16, alpha=32)
• Training Dataset: metehan777/global-seo-knowledge + Polish translations
• Training Examples: 7,880 bilingual instruction-tuning examples
• Languages: English, Polish
• Final Training Loss: 0.6887
• Final Accuracy: ~90%

Key Features

• Responds in Polish when asked in Polish
• Responds in English when asked in English
• Follows cross-language instructions (e.g., "What is SEO? Odpowiedz po polsku.")
• Covers 2,065 SEO terms and concepts across 21 categories

Training Data Composition

Type	Count	Description
Polish Q&A	2,065	Polish questions with Polish answers
English Q&A	2,065	English questions with English answers
Cross-language	2,065	Mixed language instructions
Polish (no system)	~1,000	Polish Q&A without system prompt
Instruction meta	~685	Short-form Polish instructions

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("Kelnux/Qwen3-0.6B-seo-bilingual")
tokenizer = AutoTokenizer.from_pretrained("Kelnux/Qwen3-0.6B-seo-bilingual")

# Polish question
prompt = "<|im_start|>system\nJestes ekspertem SEO.<|im_end|>\n<|im_start|>user\nCo to jest E-E-A-T w SEO?<|im_end|>\n<|im_start|>assistant\n"

# English question
prompt = "<|im_start|>system\nYou are an SEO expert.<|im_end|>\n<|im_start|>user\nWhat is backlinking in SEO?<|im_end|>\n<|im_start|>assistant\n"

GGUF Versions

GGUF quantized versions available at: Kelnux/Qwen3-0.6B-seo-bilingual-GGUF

Ollama

To run with Ollama using the GGUF version, create a Modelfile and run:

ollama create qwen3-seo-bilingual -f Modelfile
ollama run qwen3-seo-bilingual