Nemotron 3.5 Super VL

Nemotron 3.5 Super VL is an NVIDIA model in the Nemotron family, carrying the 3.5 generation label and the "Super" tier designation that NVIDIA uses to mark a mid-to-upper size point within a Nemotron release, between the smaller and larger variants of the same generation. The "VL" suffix indicates a vision-language configuration — a chat model intended to accept visual input alongside text rather than a text-only sibling. Our catalog records it as a chat model with independent serving measurements from Artificial Analysis: approximately 192.09 output tokens per second and a time to first token of around 445 ms. Those figures place it in the range where interactive, streaming chat feels responsive, with a sub-half-second wait before the first token arrives on the measured endpoint. We do not currently track a confirmed context window, parameter count, or quality benchmark suite for this entry, so those specifics should be checked against NVIDIA's own model card before committing to a deployment. In practice, models in NVIDIA's Nemotron line are published for use through NVIDIA's own inference endpoints and through third-party hosts, and tier names like Super are the main signal for where a given checkpoint sits on the size/throughput curve within its generation. If you are comparing Nemotron 3.5 Super VL against other hosted vision-capable chat models, the throughput and latency numbers above are the measurements we have; evaluate output quality on your own multimodal prompts, since we do not carry accuracy benchmarks for this entry.