{"type":"post","id":"nvidia-released-nemotron-3-5-super-vl-an-3a7957","slug":"nvidia-released-nemotron-3-5-super-vl-an-3a7957","title":"NVIDIA released Nemotron 3.5 Super VL , an open 120B MoE multimodal model with 12B active parameters.","description":"**Thesis:** Nemotron 3.5 Super VL makes the model itself a customizable multimodal component for specialized agent workflows, while the surrounding harness still determines how tools, policies, and evidence are applied.","retrieval_nugget":"**Thesis:** Nemotron 3.5 Super VL makes the model itself a customizable multimodal component for specialized agent workflows, while the surrounding harness still determines how tools, policies, and evidence are applied.","published_at":"2026-10-15","updated_at":"2026-10-15","record_date":"2026-10-15","date_kind":"published_at","topics":["agent-workflows"],"source_urls":["https://developer.nvidia.com/blog/nvidia-nemotron-3-5-super-vl-powers-specialized-agents-with-multimodal-reasoning"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:nvidia-released-nemotron-3-5-super-vl-an-3a7957","status":"published","visuals":[],"editorial_provenance":{"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:1476135e3387bad2899e0d6519774296f3bfc258637742f98361936a6c637068","reviewed_at":"2026-10-15T16:09:22.316Z","source_evidence_count":1,"verified_claim_count":4,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":4,"implication_count":1,"watch_condition_count":1,"related_record_count":0},"analysis":{"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Nemotron 3.5 Super VL makes the model itself a customizable multimodal component for specialized agent workflows, while the surrounding harness still determines how tools, policies, and evidence are applied.","observed_facts":[{"text":"NVIDIA describes a 120-billion-parameter mixture-of-experts model with 12 billion parameters active per token.","source_urls":["https://developer.nvidia.com/blog/nvidia-nemotron-3-5-super-vl-powers-specialized-agents-with-multimodal-reasoning"]},{"text":"The announcement says the model combines reasoning, instruction following, and tool use with text, image, document, and video understanding.","source_urls":["https://developer.nvidia.com/blog/nvidia-nemotron-3-5-super-vl-powers-specialized-agents-with-multimodal-reasoning"]},{"text":"NVIDIA says developers can post-train with NeMo using supervised fine-tuning, LoRA, and reinforcement learning.","source_urls":["https://developer.nvidia.com/blog/nvidia-nemotron-3-5-super-vl-powers-specialized-agents-with-multimodal-reasoning"]},{"text":"NVIDIA cites Scale AI, Ramp, and AI21 as organizations that post-trained the model for specialized tasks.","source_urls":["https://developer.nvidia.com/blog/nvidia-nemotron-3-5-super-vl-powers-specialized-agents-with-multimodal-reasoning"]}],"mechanism":"the model brings visual and text interpretation into one model, and developers can post-train it with organizational data, tools, and policies using NeMo methods including supervised fine-tuning, LoRA, and reinforcement learning.","why_now":"NVIDIA presents the release for increasingly long enterprise workflows that require agents to interpret documents, connect information, and use tools.","implications":["teams with document-heavy agent tasks can test whether one post-trained multimodal model simplifies their pipeline, but should compare end-to-end quality, latency, and cost against their current model stack."],"evidence_boundary":"the supplied primary source establishes the release, architecture, stated capabilities, customization path, and NVIDIA's examples and claims; it does not independently validate comparative performance or show results across a representative range of customer workloads.","watch_conditions":["independent evaluations on representative workflows should report task success, tool reliability, latency, and total inference cost under comparable deployment conditions."],"related_records":[],"new_branch_reason":"No prior New Runtime record was supplied for this release, so this analysis opens a distinct record on a new customizable multimodal model for agent workflows."},"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/nvidia-released-nemotron-3-5-super-vl-an-3a7957/","markdown":"https://newruntime.com/posts/nvidia-released-nemotron-3-5-super-vl-an-3a7957.md","json":"https://newruntime.com/posts/nvidia-released-nemotron-3-5-super-vl-an-3a7957.json"},"related_patterns":["harness-architecture-outlives-model-choice"],"next_reads":[{"type":"topic","path":"/topics/agent-workflows/","url":"https://newruntime.com/topics/agent-workflows/","title":"Agent Workflows","reason":"Explore the agent-workflows topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/agent-workflows.md","json_url":"https://newruntime.com/topics/agent-workflows.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/source-ledger.json","url":"https://newruntime.com/source-ledger.json","title":"Source Ledger","reason":"Inspect the public source evidence graph.","media_type":"application/json"}]}