NVIDIA Nemotron 3.5 Super VL
NVIDIA's open multimodal mixture-of-experts model (120B total, 12B active parameters) for coding, reasoning, tool use and document understanding, with image input and a 1M-token context window.
Nemotron 3.5 Super VL is a cloud text model built by NVIDIA. It is multimodal: alongside a text prompt it accepts images, then replies with generated text. Its context window handles up to 1.05M input tokens and up to 131K output tokens. Generation can be tuned with system instructions, temperature, and top-p sampling. It runs through OpenRouter using your own API key, from $0.15 per million input tokens. It also runs on your own hardware through Ollama, with no API key and no per-use fees.
- Released
- Oct 2026
- Pricing
- $0.15 / 1M in · $0.50 / 1M out
- Inputs
- Text, Images
- Context window
- 1.05M in · 131K out
- Controls
- System prompt, Temperature, Top-p
Samples
Generated with Nemotron 3.5 Super VL via OpenRouter. The same three prompts run against every text model in the catalog, shown verbatim, so the only thing that changes between two models’ answers is the model.
Explain what an API is to someone who has never written code. Do not use a restaurant or waiter analogy. Plain prose only — no headings, bullet points, or markdown formatting. Maximum 50 words.
An API is a way for two different computer programs to talk to each other. One program sends a request for information or an action, and the other responds with the answer. It lets apps share data without needing to know how the other works inside.
A two-person startup is deciding whether to build their own user authentication or use a hosted provider. Give them a clear recommendation, then the single strongest argument against your own recommendation. Plain prose only — no headings, bullet points, or markdown formatting. Maximum 150 words.
Use a hosted authentication provider. A two person team cannot afford to spend weeks designing, testing, and maintaining secure login flows, password reset logic, multi factor authentication, and compliance while also building their core product. Hosted providers handle security updates, breach monitoring, and regulatory requirements out of the box, letting the founders ship faster and avoid becoming accidental security experts. The strongest argument against this recommendation is vendor lock in and third party dependency, because handing off user identity to an external service means the startup is subject to that provider's pricing changes, outages, roadmap decisions, and privacy policies, potentially creating an expensive or impossible migration later if the relationship sours.
Write a briefing for a non-technical executive explaining why their company's spending on AI is hard to predict, and what they can do about it. Cover at least three distinct causes. Plain prose only — no headings, bullet points, or markdown formatting. Write exactly 205 words.
Executives often find artificial intelligence budgets difficult to forecast because the technology behaves more like a research project than a standard software purchase. First, model costs are usage dependent rather than fixed, so a spike in customer queries or employee adoption can multiply expenses overnight without warning. Second, the vendor landscape shifts constantly as new models, pricing tiers, and claims emerge every few months, making today’s cost estimate obsolete before the fiscal year ends. Third, hidden infrastructure and integration needs such as data cleaning, security controls, and specialized hardware frequently surface only after a pilot begins, turning a cheap prototype into an expensive enterprise rollout. To manage this uncertainty, leadership should treat AI spending as a portfolio rather than a single line item, setting aside a contingency fund that can absorb unexpected usage or vendor changes. Teams should be required to present total cost of ownership estimates that include ongoing inference fees, maintenance, and staff training, not just the initial license price. Finally, the company should establish gates between pilot and production stages, reviewing consumption data before committing to larger contracts. These steps will not eliminate volatility, but they will give leadership the visibility and cushion needed to invest in AI without risking financial surprise.
NVIDIA
NVIDIA builds the open-weight Nemotron family of language and multimodal models, tuned for enterprise reasoning, agentic workflows, and document intelligence.
developer.nvidia.com/nemotron ↗NVIDIA's 30B mixture-of-experts model with 3B active parameters, built as the execution layer for always-on agents.
NVIDIA's compact vision-language model (74B total, 9B active parameters), distilled from Nemotron 3.5 Super and tuned to fit a 24 GB NVIDIA GPU for coding, reasoning and tool use.
NVIDIA's small vision-language model (8B parameters, 4B effective), post-trained from Gemma 4 E4B for fast local conversations, image understanding and multi-turn tool use; a bundled draft model speeds up generation.
NVIDIA's multimodal Nemotron mixture-of-experts model unifying text and image understanding for enterprise Q&A, summarization, and document intelligence.