ModelsMulti-source

腾讯发布 Hunyuan 4 Flash,120亿激活参数、解码速度超150 tokens/秒

Tencent launches lower-cost Hy4 Flash AI model

Why we picked it

腾讯这次发的 Hunyuan 4 Flash 是轻量版,120 亿激活参数、每秒能跑 150+ tokens,调用成本更低。做高并发应用或者想省钱接 API 的开发者可以对比一下豆包和 Qwen 的快速版再选。

腾讯推出 Hunyuan 4 Flash,定位为低成本、高速度的轻量级模型。该模型拥有 120 亿激活参数,解码速度超过每秒 150 tokens,主打推理吞吐与调用成本之间的平衡。

Body

与标准版 Hunyuan 4 相比,Flash 版通过缩减激活参数规模换取更快的响应速度,适合高并发、低延迟场景,例如客服对话、内容审核和搜索增强等应用。按激活参数计算,它属于 MoE 架构下的中小规模模型,部署成本明显低于旗舰模型。

国内主要厂商近月密集发布轻量模型线,包括阿里 Qwen 系列、字节豆包的快速版模型,Flash 类产品已成为各家争夺企业 API 客户的主力。腾讯此次以速度和成本作为卖点,直接面向价格敏感的批量调用需求方。

Original · Tech in Asia

Tencent launches lower-cost Hy4 Flash AI model

Tencent also said the model has 12 billion active parameters and decoding speeds above 150 tokens per second.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Models" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Hunyuan, Tencent, open-source models.

Source attribution: curated from Tech in Asia (@Diya Lal). Cross-source corroboration: 8 independent sources have covered this development (including pandaily, OpenRouter, Blognone), confirming industry-wide signal strength.