← Back
NVIDIA Blog

From Training to Production, NVIDIA and CoreWeave Close the Loop on Agentic AI

•6 min read•
#agents#deployment#nvidia
Level:Advanced
For:AI Engineers
✦TL;DR

NVIDIA’s Vera Rubin NVL72 systems, now available on CoreWeave Cloud, deliver up to 4.8× higher token throughput for agentic coding workloads compared with the prior‑generation GB200 NVL72. Cognition, the AI lab behind the Devin software‑engineer agent, benchmarked this gain on real‑world software‑engineering tasks, showing faster real‑time code generation and more responsive multistep reasoning. The rollout leverages CoreWeave’s end‑to‑end stack—including Kubernetes Service, SUNK, Mission Control, Sandboxes, and Inference—to bring the full NVIDIA AI factory to production agents. Engineers can therefore achieve higher throughput and lower latency for agentic AI without redesigning their pipelines.

⚡ Key Takeaways

  • 4.8× increase in total token throughput for SWE‑2 inference workloads on Vera Rubin NVL72 vs. GB200 NVL72.
  • CoreWeave provides a full‑stack deployment environment (Kubernetes Service, SUNK, Mission Control, Sandboxes, Inference) for NVIDIA hardware.
  • Performance boost translates to faster real‑time code generation and lower cost‑per‑token for multistep agentic workloads.
  • Production clusters can be stood up in days via CoreWeave’s managed services, enabling rapid access to the new hardware.
  • The Vera Rubin CPU for AI agents is announced but not yet available; early‑access customers must use the NVL72 GPU platform first.
💡 Why It Matters

The near‑5× token‑throughput uplift lets teams ship agentic coding assistants at scale with lower latency and reduced token costs, directly impacting product responsiveness and operating expense.

✅ Practical Steps

  1. Deploy a Vera Rubin NVL72 cluster on CoreWeave Cloud using CoreWeave Kubernetes Service.
  2. Route your agentic inference workloads (e.g., Devin) through CoreWeave Mission Control and CoreWeave Inference to leverage the 4.8× token‑throughput gain.
  3. Use CoreWeave Sandboxes for isolated agentic sandbox startups, benefiting from the reported >3× faster startup times.

Want the full story? Read the original article.

Read on NVIDIA Blog ↗

More like this

Introducing workload recovery and expanded infrastructure checks on Amazon SageMaker HyperPod

AWS ML Blog•#deployment

Introducing Unity Gateway Insights

Databricks Blog•#deployment

Introducing Gemini 4 Neon

Google DeepMind•#llm

DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency

Hugging Face Blog•#llm

EXPLORE AI NEWS

Daily hand-picked stories on LLMs, RAG, agents and production AI — curated for engineers who ship.

BROWSE NEWS

GET THE WEEKLY DIGEST

Join engineers getting the Monday signal-over-noise AI breakdown. No spam, unsubscribe anytime.

LEARN AI ENGINEERING

Curated courses, research papers, repos and tutorials built for engineers leveling up in AI.

START LEARNING