[KongchangAI]
Model腾讯混元 / Hunyuan

Hunyuan

A large language model developed by Tencent, supporting multiple AI application scenarios including AI-assisted programming.

Timeline (last 90 days)

Oct 19

Hy-Memory-30B-A3B uses the Hunyuan sparse expert (MoE) architecture, with about 30 billion total parameters and only about 3 billion parameters activated per forward pass.

Unverified50%
Oct 17

The "Mini Program Growth Plan" of the WeChat Mini Program development competition "Symbiosis with AI" runs from January 1, 2026 to December 31, 2026, and participants can obtain free quota and cloud service resources, with packages including cloud development resources and free Hunyuan tokens.

Unverified50%
Oct 17

Domestic large models such as MiniMax, Kimi, Tongyi Qianwen, and Hunyuan can all serve as backend options for Agent tools.

Unverified50%
Oct 5

Hunyuan was rated "number one among domestic models" with an overall ranking of fourth.

Expired50%
Oct 5

The domestic AI models competing in this episode include four: Hunyuan, GLM, DeepSeek, and Xiaomi.

Expired50%
Sep 17

MoE models are well-suited for high-frequency agentic workloads, such as applications involving a large number of tool calls.

Unverified60%
Aug 31

This compression reduced the Hunyuan model size by nearly 87%.

Unverified50%
Aug 31

The compressed Hunyuan version is a preview release, and its stability and completeness remain to be observed.

Unverified50%

All Facts (6)

Source Articles