Tencent has made its Hy3 large language model available for global use. The model, which was previously known as Hunyuan, utilizes a hybrid Mixture-of-Experts (MoE) architecture.
Hy3 features 295 billion parameters.
Supported context length for Hy3
The model supports a context length of 256,000 tokens. Users can access Hy3 through Tencent Cloud TokenHub, Tencent Design Miora, and WorkBuddy.
Internal tests by Tencent show that WorkBuddy updated with Hy3 achieves a success rate of over 90% in tasks and reduced the average task completion time by 34% compared to the previous model.
Licensing and performance
Hy3 is released under the Apache 2.0 license for commercial use. Within its first week of release, Hy3 reached the top of OpenRouter's large language model usage rankings.
Starts at $0.1288 per 1 million input tokens and $0.5336 per 1 million output tokens
