Hot in AI

Hottest right now

AIdeveloper.nvidia.com

Nvidia releases Nemotron 3.5 Super VL, an open 120B multimodal model for agents

Nvidia has released Nemotron 3.5 Super VL, a mixture-of-experts model with 120 billion parameters, of which 12 billion are active for each token. It works with text, images, documents and video, and is aimed at AI agents. Nvidia published base, post-trained and NVFP4-quantized checkpoints on Hugging Face and ModelScope, along with training recipes, data and reinforcement-learning environments. The model is also available through OpenRouter and as an NVIDIA NIM. Nvidia says it delivers up to five times the throughput of other open models its size on agentic tasks on a single GPU; that figure comes from its own tests.

  1. AImi5.gov.uk

    MI5 launches a centre to assess AI threats to UK national security

    MI5 has set up an AI Threat Assessment Centre (AITAC), Director General Sir Ken McCallum said in his annual threat update on 13 October. The centre is based in MI5 and brings in expertise from the AI Security Institute, the National Crime Agency and the National Cyber Security Centre. It produces all-source assessments of AI-related national security threats for government departments. In the same speech, McCallum said AI is already intensifying the contest between cyber attackers and defenders, and that terrorists have tried to use it to improve their reconnaissance and attack methods. The misuse he would weigh most heavily, though it has not happened yet, is AI changing the stakes for existing or new bioweapons.