Tencent will release Hunyuan Hy4 and Hy4 flash, upgrading its flagship large model with visual understanding for the first time and offering a cost-efficient variant, Dowson Tong said on October 29 at the 2026 Tencent Global Digital Ecosystem Summit1,2.
Tong, Tencent's senior executive vice president and CEO of the Cloud and Smart Industries Group, said the full Hy4 model improves performance on tasks requiring sustained reasoning and execution, including real-world productivity workflows. Compared with the Hy4 preview released and open-sourced in August, the production model delivers comprehensive capability upgrades, strengthens handling of long-horizon tasks, and significantly improves chain-of-thought efficiency.
The Hy4 preview has 770 billion total parameters with 49 billion activated during inference and supports a context window of one million tokens. Tong said Tencent is developing its models and products together to improve their performance in practical applications.
Hy4 flash, positioned for cost efficiency, will retain visual understanding and AI agent capabilities. No release date was specified for either model beyond the near-term framing.
On the same day, Tencent and Visa signed a memorandum of understanding to establish a cooperation framework in cross-border payments, agentic commerce, enterprise AI, and corporate globalization. The companies plan to deepen cross-border remittance cooperation through Visa Direct.
The Hy4 announcement follows Tencent-Hunyuan's October 27 open-source release of Hy-SimulMT, a 30B-parameter mixture-of-experts model for simultaneous speech-to-speech interpretation between Chinese and English[3].
ANALYSIS Adding visual understanding to a 770B-parameter architecture while simultaneously shipping a cost-optimized flash variant pairs a full-capability flagship with a lighter model aimed at high-volume deployment.