NVIDIA's Nemotron 3.5 Super VL is free on OpenRouter: 1,000,000-token context with image, document, and video input, at $0 input and output.
The open multimodal mixture-of-experts model (120B total, 12B active parameters) accepts text, images, GUI screenshots, and video for document understanding and agent work; NVIDIA published the official weights on Hugging Face this week.
The catch: free-tier OpenRouter models run under shared rate limits that vary by model and load, and heavy video input burns the quota fast.
How to claim: 1) Create an OpenRouter account → 2) Call the model nvidia/nemotron-3.5-super-vl:free via API or on the model page → 3) Keep usage within the free quota.
Deadline: Ongoing.
Link: Nemotron 3.5 Super VL on OpenRouter
NVIDIA: 1M-Context Nemotron 3.5 Super VL Free on OpenRouter
1M-token context, $0 input and output, image and video input
Catch: shared OpenRouter free-tier rate limits, no announced end date
Ongoing; claim with an OpenRouter account and the :free model id
Oct 17, 2026, 2:08 AM · NVIDIA
Offers mentioned
- NVIDIA NIM Free Inference Endpoints Verified Oct 12, 2026
Access free inference endpoints and test leading AI models directly through NVIDIA NIM APIs.
Related guides
- Free AI APIs With No Credit Card in 2026
15 free AI APIs that need no credit card: LLM chat, embeddings, image and speech, with the free limits and what you give up instead (data, phone, use).
- Free DeepSeek API in 2026: What Is Actually Free
DeepSeek’s own API is pay-as-you-go, half price off-peak. Seven providers run DeepSeek V4 Flash models free, with real limits, model IDs and the catch.
- How to Apply for AI Startup Credits
Who qualifies for startup credits from Anthropic, OpenAI, AWS, Together AI, fal and GitHub, what each application asks for, and the catches.
Source: NVIDIA · More info