Nvidia DeepSeek-V4.1-Flash-1017-NVFP4 tops Hugging Face likes list
Esta notícia ainda só está disponível em inglês.
Nvidia's DeepSeek-V4.1-Flash-1017-NVFP4 emerged as the most liked model on Hugging Face on 2026-11-02, recording 105 likes and 1,473 downloads.
Popularity metrics
The model, tagged for image-text-to-text pipelines, led the daily trending list with 105 likes and 1,473 downloads, indicating strong community interest on its release day.
Model specifications
According to nvidia, the model supports contexts of up to one million tokens and contains 664.9B total parameters, of which 552B form the backbone. It activates 8B parameters during prefill and 16B during decode, and includes a 196B Engram conditional memory. According to nvidia, the model accepts text and image as input types, outputs text in string format, and is quantized with nvidia-modelopt v0.48.0rc0. The checkpoint is approximately 400 GiB in size, and the model is compatible with NVIDIA Blackwell hardware, runs on Linux, and is ready for commercial or non-commercial use. It was released on Hugging Face on 10/30/2026.
Pontos principais
- The DeepSeek-V4.1-Flash-1017-NVFP4 model combines massive scale with versatile multimodal input.
- Its release generated the highest engagement among new Hugging Face models on the day.
- Nvidia positions the model for broad commercial deployment on Blackwell GPUs and Linux environments.
Anúncio oficial
Publicado por nvidia. Leia na fonte.