ModelsNov 18, 2026 · 23:528 sources

Tencent Releases Open Weights for Hy4 Flash Multimodal AI Model

From marketbrief.now

Image: marketbrief.now
AI;DR

Tencent released the open weights for its Hy4 Flash multimodal AI model, including FP8 and Base versions. The model uses a mixture-of-experts architecture with 224B total parameters and 12B active parameters per token, supporting a 1M token context window. The weights are distributed under the Tencent Hy License.

The release coincides with day-0 support from the vLLM and SGLang runtimes. Tencent claims Hy4 Flash slightly outperforms its larger 770B parameter Hy4 Preview on productivity tasks while costing 1/13 as much per task. Following a surge in demand for the model, Tencent's WorkBuddy platform expanded its inference cluster tenfold.

Read the sourceSign in to suggest

Sources

+ 4 more
  1. SUPPORT@sino_market“WorkBuddy has expanded the Hy4 inference cluster tenfold and raised free-usage quotas”x.com
  2. SUPPORT@himanshustwts“what is this pricing? how are they able to serve it this cheap?”x.com
  3. SUPPORT@martinvars“On those tasks it slightly beats the much bigger Hy4 Preview (770 billion parameters) at 1/13 of the cost per task”x.com
  4. SUPPORT@teortaxestex“Been a while since we got a small MoE base of this caliber”x.com
#open-sourceMarkdown
Content history (1)
  • 2026-11-18 · Summary · vi · Replaced stored text
    Tencent vừa công bố phát hành open-weight cho Hy4 Flash, một mô hình AI đa phương thức củ…
    Tencent đã phát hành open-weight cho mô hình AI đa phương thức Hy4 Flash, bao gồm các phi…

One ranked edition, every morning

The stories that matter today. Free, cancel anytime.

Telegram @aidr_today