JibayAi/Jibay_2
JibayAi/Jibay_2 is a 2 billion parameter, open-source language model built upon the Qwen 3 family, designed to be simple, trainable, and customizable. Released by JibayAi, it features a 32,768 token context length and is optimized for fine-tuning, on-device AI, and rapid prototyping without requiring extensive computational resources. This model serves as an efficient alternative for developers needing a controllable and adaptable LLM for various domain-specific applications.
Loading preview...
Jibay 2: A Customizable 2B Parameter LLM
Jibay 2, developed by JibayAi, is a lightweight, open-source language model based on the Qwen 3 architecture. Publicly available since May/June 2026, it emphasizes simplicity, trainability, and customization, making it an efficient choice for developers and researchers.
Key Capabilities & Specifications
- Architecture: Built on Qwen 3, featuring 2 billion active parameters.
- Context Length: Supports a substantial 32,768 tokens (input + output).
- Customization: Designed for easy retraining, extension, and adaptation to any field or language.
- No MoE: Does not use Mixture of Experts, simplifying understanding, debugging, and fine-tuning.
- Performance: Achieves ~56.9% on MMLU, ~70.0% on GSM8K, and ~68.0% on HumanEval for Python code generation.
- Recommended Usage: GGUF quantized version is strongly recommended for optimal performance, especially on consumer hardware.
Good For
- Fine-tuning: Ideal for fine-tuning on domain-specific datasets (e.g., medical, legal, niche code).
- Edge AI: Suitable for on-device or edge AI applications due to its lightweight nature.
- Research & Education: Excellent for academic research and educational purposes.
- Agentic Workflows: Capable of supporting lightweight agentic workflows with function calling capabilities.
- Rapid Prototyping: Facilitates quick experimentation and prototyping without major infrastructure needs.