<ul data-eligibleForWebStory="true"><li>Tencent has released an open-source large language model called Hunyuan-A13B, based on a sparse MoE architecture with 13 billion active parameters.</li><li>Hunyuan-A13B features Grouped Query Attention, 256K context length, and dual-mode reasoning for fast and slow thinking to balance computational cost and performance.</li><li>The model achieves top-tier performance in agentic benchmarks, excelling in tool-calling and long-context scenarios, surpassing larger models in logical reasoning and agent tasks.</li><li>Hunyuan-A13B, optimized for efficient deployment, supports inference optimization formats and frameworks, making it suitable for real-time applications and long-context tasks.</li></ul>

Tencent Open Sources Hunyuan-A13B: A 13B Active Parameter MoE Model with Dual-Mode Reasoning and 256K Context

Discover more