NVIDIA Releases Nemotron 3.5 and OpenAI Expands Lab Access
NVIDIA launched open models for continuous agents. Meanwhile, OpenAI introduced free access to its life-sciences model for academic labs.
NVIDIA Releases Nemotron 3.5 Super VL for Autonomous Agents
On October 15, 2026, NVIDIA announced the open-weight Nemotron 3.5 Super VL model. It features 120 billion total parameters while activating only 12 billion parameters per token. The system natively handles text, images, documents, and video without requiring extra vision plugins. NVIDIA reported up to 5x higher task throughput on a single graphics processor compared to similar open models. Alongside it, the company released Nemotron 3.5 Bolt, which fits within 24 gigabytes of memory for local desktop agent workflows.
What it means: Organizations can run full multimodal agents on internal machines instead of relying on external cloud providers. This architecture lowers processing costs while keeping private company data strictly on premise.
Example: An automated office assistant reviews incoming invoice scans, checks purchase orders, and logs financial entries on a local workstation.
- Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models | NVIDIA Blog · 15 October 2026
- NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal Reasoning | NVIDIA Technical Blog · 15 October 2026Background
- AGI HUNT · AI News Daily 2026-10-16 — Today's AI Highlights · 16 October 2026Background
- [AI Frontier Daily] 2026-10-16 | Stillflood · 16 October 2026Background
- Projects - NVIDIA Nemotron · 15 October 2026Background
Also worth knowing
OpenAI Unveils Rosalind Academic Plan for Life Sciences
On October 15, 2026, OpenAI introduced the Rosalind Academic Plan for qualified life-science laboratories. Eligible university labs receive up to 30 free seats for 12 months. The specialized GPT-Rosalind model assists researchers with genomics, medicinal chemistry, and experimental planning under strict vetting.
What it means: Academic research teams can run specialized frontier scientific workflows without paying expensive enterprise software licenses.
Example: A university research group analyzes protein structures and designs preliminary drug trials before scheduling physical experiments.
- Bringing GPT-Rosalind to academic research labs | OpenAI · 15 October 2026
- Daily AI brief: today’s AI news, research and releases · Kush Ahuja · 16 October 2026Background