Tencent (Chinese tech company)

téngxùn

腾讯

to release / to announce

fābù

发布

open-source

kāiyuán

开源

large language model / LLM

dà móxíng

大模型

Hunyuan 3 (Tencent AI model)

hùnyuán sān

混元3

preview version / beta release

yùlǎn bǎn

预览版

。

New Words

News in English

Tencent Releases Open-Source Large Language Model Hunyuan 3 Preview

Tencent Holdings released Hunyuan 3 Preview on April 24, a large language model using a mixture-of-experts architecture with 295 billion total parameters and a 256,000-token context window, made available as open-source.

On April 24, Tencent Holdings officially released a preview version of Hunyuan 3, marking the first major AI model launched by the company's new management team since they took office.

Hunyuan 3 employs a mixture-of-experts architecture with a total of 295 billion parameters. During each inference operation, only approximately 21 billion parameters are activated, making it efficient for processing tasks.

The model supports ultra-long context windows of up to 256,000 tokens, making it suitable for processing complex long-document tasks.

Tencent released Hunyuan 3 in fully open-source form, allowing developers to freely access the model weights on platforms such as Hugging Face.

According to benchmark data released by Tencent, Hunyuan 3 demonstrates excellent performance across multiple industry benchmarks, particularly excelling in code generation and multi-modal understanding.

This release marks Tencent's sustained commitment to increasing investment in the open-source AI competitive landscape, engaging in direct competition with multiple domestic and international tech giants.

More in Science & Technology