Chat2Find/chat2find-instruct-v1

VISIONConcurrency Cost:1Model Size:4.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:May 30, 2026License:mitArchitecture:Transformer0.0K Open Weights Cold

Chat2Find/chat2find-instruct-v1 is a 4.5 billion parameter trilingual instruction-tuned model developed by Chat2Find, based on a continued pre-trained Qwen3.5-7B architecture. Optimized for chain-of-thought (CoT) reasoning and robust agentic tool calling, it supports complex instruction-following in Sinhala, Tamil, and English. With a native context window of 262,144 tokens, it excels at solving complex mathematical, logical, and agent-driven workflows.

Loading preview...

Chat2Find-Instruct-v1: Trilingual Reasoning & Agentic Model

Chat2Find-Instruct-v1 is a 4.5 billion parameter trilingual model, built by Chat2Find, specifically optimized for advanced reasoning and agentic capabilities. Based on the robust Qwen3.5-7B architecture and fine-tuned with high-quality datasets, it processes queries with deep chain-of-thought (CoT) reasoning and robust agentic tool calling.

Key Capabilities

  • Deep Chain-of-Thought (CoT) Reasoning: Automatically processes queries within <reasoning> tags before generating a final output in <answer> tags.
  • Robust Agentic Tool Calling: Natively parses complex system prompts, invokes external APIs/functions, and digests responses to fulfill user objectives.
  • Premium Trilingual Support: Seamlessly understands and switches between English, Sinhala, and Tamil.
  • Lightweight & High Efficiency: Fine-tuned using Unsloth for optimized performance and memory footprint.

Training Details

The model was fine-tuned on the Chat2Find Unified Reasoning & Tool Dataset (279,260 records), with a significant focus on Tamil (45%), Sinhala (36%), and English (18%) data. The dataset included 78.4% general SFT tasks and 21.6% multi-turn agentic chat.

When to Use This Model

This model is ideal for applications requiring:

  • Complex problem-solving that benefits from explicit reasoning steps.
  • Integration with external tools and APIs for dynamic, real-time information.
  • Multilingual interactions, particularly in Sinhala, Tamil, and English contexts.
  • Efficient deployment due to its optimized architecture and training.