Tencent HunyuanLaunchAI ModelsNovember 18, 2026

Tencent open-weights Hy4 flash, a 224B multimodal model

Read original source →huggingface.co

Hy4 flash activates 12B of its 224B parameters, supports a 1M-token context and image inputs, and runs at roughly 200 tokens/sec. Tencent says it processed over 50 trillion tokens on day one, its best Hunyuan first-day usage, and priced API access at $0.07/M input and $0.3/M output tokens.

How this story unfolded

4 days · 6 reports · 8 community posts · 14 of 19 shown

  1. Nov 16
  2. Nov 18
  3. Nov 19
  4. Nov 20

More stories today

Open the live feed