Nemotron 3 Ultra Powers Agent Workflows
LangChain · x · 2026-07-16
LangChain shared a new video from @itsao detailing how to run NVIDIA Nemotron 3 Ultra on Baseten and integrate it into a Deep Agents Code workflow.
Highlighted capabilities include:
- 550B parameters, reaching speeds of 300 tokens/s
- Endpoint Agent support for skills, sub-agents, MCP
- First-class tracing provided by LangSmith
Ultimately, this is an engineering-focused demonstration of an agent/coding workflow rather than mere model promotion.
Related event: LangChain Demonstrates Nemotron 3 Ultra Agent Workflow(4 posts)→
More from coding & agent
- WanGP adds one-click AMD GPU support for Tencent's Hy-Nomas world model, runs on 8GB VRAM — cocktailpeanut · 2026-10-27
- He benchmarks new models by having them build their own copilot — $28 of compute — bucolucas · 2026-10-27
- Founder generates 46 app promo videos in 4 days with Claude, skipping $11,500 in UGC fees — PrajwalTomar_ · 2026-10-27
- Should agents be allowed to change plans mid-task? Reddit debates the drift boundary — wistfulcracker84 · 2026-10-27
- HeyGen + Canva MCP turns slide speaker notes into an avatar presentation video — HeyGen · 2026-10-27
- Local beats cloud: Qwen 3.8 FN handles private student data as well as frontier models — dr_lm · 2026-10-27