Nvidia Targets AI Giants with Nemotron 3.5 Super VL

Introduction

Nvidia has introduced Nemotron 3.5 Super VL, a vision-language model that targets efficiency in agentic tasks.

Model Architecture

The model pairs two components: NemotronH_Omni_Reasoning_V3 and RADIO v2.5-H encoder.

Benchmark Claims

The model claims to outperform Claude Opus 5 and GPT-6.1 Sol on specialized benchmarks, but the claims lack independent verification.

Open-Weights Approach

Nvidia's strategy is centered around open weights, allowing companies to download and customize the model for their needs.

Conclusion

The Nemotron track record has delivered up to 2.2 times the throughput of GPT-OSS-120B while holding competitive accuracy across various benchmarks.

Future Developments

GTC Berlin is scheduled for October 20-22, 2026, where the model will be discussed further.