Tencent launches Hy3, a 295B MoE model focused on agent performance

Tencent has officially launched its Hunyuan Hy3, a 295 billion parameter Mixture-of-Experts (MoE) model with 21 billion active parameters and a 256K context window. The model is licensed under Apache 2.0 and emphasizes practical agent performance, showing significant improvements in task resolution and real-world applications. While not the largest model, Hy3 is positioned as Tencent's most pragmatic AI offering, integrated into over nine of its products and priced competitively. AI

IMPACT This pragmatic approach to AI development and Tencent's vast distribution network could significantly impact enterprise adoption, especially in the Asian market.

RANK_REASON Frontier-lab model release with system card [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on dev.to — LLM tag →

• Apache 2.0
• DeepSeek V4 Pro
• GLM-5.2
• GPT-5.5
• Hunyuan Hy3
• Mixture-of-Experts
• Qwen 3.7 Max
• Tencent
• WeChat
• WorkBuddy

• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →