Upgraded to a MOE structure with a context window of 256k, leading many open-source models in various NLP, coding, mathematics, and industry benchmarks.

Utilizes a superior routing strategy while alleviating issues of load balancing and expert convergence. For long texts, the needle-in-a-haystack metric reaches 99.9%. MOE-32K offers a relatively higher cost-performance ratio, balancing effectiveness and price while enabling processing of long text inputs.

Utilizes a superior routing strategy while alleviating issues of load balancing and expert convergence. For long texts, the needle-in-a-haystack metric reaches 99.9%. MOE-256K further breaks through in length and effectiveness, greatly expanding the input length capacity.

The preview version of the next-generation Hunyuan large language model, featuring a brand-new mixed expert model (MoE) structure, which offers faster inference efficiency and stronger performance compared to Hunyuan Pro.

A trillion-parameter scale MOE-32K long text model. Achieves absolute leading levels across various benchmarks, capable of handling complex instructions and reasoning, with advanced mathematical abilities, supporting function calls, and optimized for applications in multilingual translation, finance, law, and healthcare.

The latest multimodal model from Hunyuan, supporting image + text input to generate textual content.

The latest code generation model from Hunyuan, trained on a base model with 200B high-quality code data, iteratively trained for six months with high-quality SFT data, increasing the context window length to 8K. It ranks among the top in automatic evaluation metrics for code generation across five major programming languages, and performs in the first tier for comprehensive human quality assessments across ten aspects of coding tasks.

The latest MOE architecture FunctionCall model from Hunyuan, trained on high-quality FunctionCall data, with a context window of 32K, leading in multiple dimensions of evaluation metrics.

The latest role-playing model from Hunyuan, fine-tuned and trained by Hunyuan's official team, based on the Hunyuan model combined with role-playing scenario datasets for enhanced foundational performance in role-playing contexts.