Nemotron 3.5 Super VL Is Free on OpenCode

NVIDIA’s Nemotron 3.5 Super VL is now free on OpenCode (s'ouvre dans un nouvel onglet). It has 120B total parameters, with 12B active per token, and can work with text, images, and video. NVIDIA built it for tool use and long-running, multistep agent tasks. NVIDIA’s announcement (s'ouvre dans un nouvel onglet)

Unlike a regular chat model, Super VL is designed to follow instructions, use tools, and understand visual inputs in one workflow. NVIDIA reports strong speed and accuracy, but those are vendor claims—not proof it will outperform your current coding model on your own repo.

OpenCode lists input and output tokens as free, but access is temporary and marked as a trial. NVIDIA says session data may be logged to improve its products, so don’t send private or confidential code. OpenCode model details (s'ouvre dans un nouvel onglet)

Worth trying? Yes, for free experiments with repo exploration, visual tasks, and tool calling. Compare it with DeepSeek V4.1 Flash Free (s'ouvre dans un nouvel onglet) on the same task. Keep whichever completes work reliably; don’t depend on a temporary free endpoint.