Tencent Holdings Limited (00700) has officially launched its Hunyuan Hy3 large language model. This model represents the formal iteration of the Hy3preview version released in April, maintaining a Mixture of Experts (MoE) architecture with 295 billion total parameters and 21 billion activated parameters, alongside a 256K context length.

Analyst Perspective and Valuation

The firm maintains its "Overweight" rating on Tencent with a target price of HK$632. The official launch of Hy3 validates Tencent's execution capabilities in enhancing model reliability through post-training, which is expected to prompt the market to reassess the implied value of Tencent's AI assets.

Cost Efficiency Advantages

On the cost front, the 21 billion activated parameters enable Hy3 to operate at a significantly lower computational cost compared to flagship models of similar capability. Its API input price has been reduced to 1 RMB. Data from WorkBuddy indicates that in practical business applications, token consumption is approximately 47% to 49% lower than with GLM-5.2. This directly reduces the marginal cost of Tencent Cloud's AI services and provides a manageable path to scaling for large-scale consumer-facing applications such as Yuanbao, ima, and WeChat Read.

Revenue Growth Drivers

On the revenue side, the substantial improvement in model capability is expected to drive increased usage of Tencent Cloud's large model API. Furthermore, metrics such as a 37.6% improvement in programming task success rates for QQ Browser and a 14.1% increase in label accuracy for WeChat Read reflect the model's positive impact on specific product experiences, indirectly supporting user engagement time and commercial conversion.

Ecosystem Expansion and Internal Productivity

Regarding the ecosystem, Hy3 deployment has already covered key scenarios including office productivity, AI agents, gaming, customer service, and reading. The completion of the full pipeline from infrastructure rebuild to product-level models between late January and July signifies that Tencent's AI investments are transitioning from expense items to internal productivity tools. The resulting boost to R&D efficiency and product iteration speed is anticipated to become evident in subsequent quarters.