Updated

Hy4 flash

Tencent released Hy4 flash, a multimodal model with a context length of up to one million tokens and 12 billion active parameters. Tencent says it is designed for everyday productivity tasks with lower inference cost than Hy4.

By No Code MBA · Facts checked

Provider
Tencent
Announcement date
Status at last check
Generally available

What changed

Hy4 flash adds a lower-cost member to Tencent's Hy model family, with 224 billion total parameters, 12 billion active parameters, multimodal input, and up to a one-million-token context. Tencent describes hybrid attention, LatentMoE, and Engram features, with a smaller KV cache than Hy3 and Hy4.

Access and availability

Available through Tencent Cloud TokenHub and Tencent products including WorkBuddy, Yuanbao, WeCom, QQ, ima, and Marvis. Tencent says OpenRouter availability is planned, not yet available.

Official access documentation

What you can build

Compare its cost and output quality with Hy4 on representative document, presentation, and research tasks before using it for longer software-engineering work.

Limitations

Tencent recommends Hy4 rather than Hy4 flash for complex, long-horizon software engineering. Performance and cost comparisons in the announcement are Tencent-reported evaluations.

Sources and release history