Qwen3 32B

About this model

Qwen3 32B is a base foundation model with 32 billion parameters and 262K native context, designed for fine-tuning and custom adaptations. Pre-trained on diverse multilingual data covering 77.5% of languages, providing strong general capabilities across text understanding, code, mathematics, and reasoning. Serves as the foundation for specialized models and custom fine-tuning projects requiring a powerful mid-sized base. Ideal starting point for domain-specific adaptations and research applications.

Technical specifications

Capabilities

Input modalities

Output modalities

Reasoning

Hybrid Default off

Knowledge horizon