LLM
Nemotron 3.5 Super VL
Nemotron 3.5 Super VLAn open 120B-A12B MoE model built for agent workflows across text, images, and video.
NVIDIA Nemotron 3.5 Super VL brings visual understanding to the Nemotron Super family. This 120B-A12B mixture-of-experts model processes text, images, and video with a 1M-token context window, helping agents reason across documents and visual evidence, then use tools to complete multi-step tasks. It’s designed for workflows such as document intelligence, customer service, enterprise productivity, and web research.
Run it on Baseten Dedicated Inference or Model APIs, then customize it with your data, tools, and policies using supervised fine-tuning or reinforcement learning in Baseten Loops.
Input
1from openai import OpenAI
2
3client = OpenAI(
4 api_key="BASETEN_API_KEY",
5 base_url="https://inference.baseten.co/v1"
6)
7
8# NVIDIA recommends temperature=1.0 and top_p=0.95 for all tasks.
9# Toggle reasoning on/off via chat_template_kwargs.enable_thinking.
10response = client.chat.completions.create(
11 model="nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B",
12 messages=[{"role": "user", "content": "Hello, what can you do?"}],
13 max_tokens=16000,
14 temperature=1.0,
15 top_p=0.95,
16)
17print(response.choices[0].message.content)
