Trendyol/Trendyol-LLM-8B-T1

Trendyol LLM-8B-T1: Turkish E-commerce Reasoning

Trendyol LLM-8B-T1 is an 8-billion-parameter chat model developed by Trendyol, leveraging the Qwen 3-8B architecture. Its primary distinction lies in its specialized training on extensive Turkish e-commerce datasets, making it highly proficient in advanced Turkish reasoning tasks. Crucially, it maintains the robust English capabilities of its base model, offering strong performance in both languages.

Key Capabilities & Features

• Turkish Reasoning: Provides robust chain-of-thought reasoning across various domains in Turkish.
• English Reasoning: Preserves strong reasoning capabilities in English, including chain-of-thought examples.
• Dual Operation Modes: Features /think for explicit internal reasoning (default) and /no_think for concise answers.
• Multitask Performance: Excels out-of-the-box in instruction following, summarization, paraphrasing, coding tasks, text/review classification (sentiment, category), and attribute/key-value extraction for catalogue enrichment.
• E-commerce Tuned: Incorporates domain-specific vocabulary for sectors like fashion and electronics.
• Extended Context: Supports a native context length of up to 32,000 tokens.
• Apache-2.0 License: Available for both commercial and research use.

Use Cases & Considerations

This model is particularly well-suited for applications requiring sophisticated Turkish language understanding and generation, especially within the e-commerce domain. Its dual-language proficiency makes it versatile for environments serving both Turkish and English-speaking users. Developers should be aware of its limitations, including potential for generating false information and biases inherited from its training data, and are encouraged to implement human oversight and application-specific safety testing.