tencent/Hy4-flash
image-text-to-text · ♥ 591 · ↓ 6.6K
- Taskimage-text-to-text
- Downloads3949
- Likes558
- Licenseother
image-text-to-text · ♥ 591 · ↓ 6.6K
Hy4 flash is our most cost-effective Hy model to date, featuring 220B total parameters (including 64B Engram parameters) with just 12B active parameters per token. It natively accepts text and image inputs, generates text, and supports a context window of up to 1M tokens.
Its 42-layer language backbone consists of one dense feed-forward layer and 41 MoE layers, with identity Hyper-Connections (iHC) supporting inter-layer information flow. A vision encoder (adapted from SigLIP2-So400m-Patch16-Res512) is connected to the backbone through an MLP projector. In addition to the backbone, a built-in 5-layer drafter, DFly2 (0.85B parameters), reuses the target model's KV cache and sparse…
Excerpt from the model card on Hugging Face ↗.
tencent/Hy4-flash was first seen on Hugging Face · Trending models on Nov 18, 2026. Since then it has appeared in 106 checks on 5 different days. Its best position was #1 of 30, reached on Nov 20, 07:45 UTC. It is #4 right now, holding steady at #4 over the last 6 checks.
Position on Hugging Face · Trending models at every check (lower is better).
Best rank per day (UTC).
🔥 as reported by Hugging Face.
Archived daily, weekly and monthly lists that include this entry.
Yes. As of Nov 22, 16:45 UTC it is #4, holding steady at #4 over the last 6 checks.
#1 on Hugging Face · Trending models, on Nov 20, 07:45 UTC. Trendoday checks this list every 60 minutes.
It was first seen on Nov 18, 2026 and has been on the list on 5 days (106 checks) since.
It is tagged image-text-to-text with transformers on Hugging Face and has 3949 downloads under the other license.