NVIDIA published the first on-silicon benchmark metrics for its Vera Rubin architecture on long-horizon agent workloads, demonstrating up to 30x throughput per megawatt and 35x lower token cost versus GB300 NVL72.

Key Takeaways

  • ✓Inference benchmarks rebased from short chat to multi-step agent coding workloads with deep context expansion;
  • ✓Benchmarked against DeepSeek V4 Pro on SemiAnalysis AgentX real coding traces as the reference workload;
  • ✓Delivers up to 30x throughput per megawatt and 35x lower token cost compared to GB300 NVL72.
⚡

Key Decision Metrics at a Glance

TL;DR Verdict
🏆 Benchmark RankRank
Benchmarked against DeepSeek V4 Pro on SemiAnalysis AgentX real coding traces as the reference workload;
Full Leaderboard→
💰 Cost & PlansCompare
DeepSeek-V4-Pro · in: $0.66 / out: $1.98
Compare Plans↗
🛠️ Ecosystem & IDERules
Cursor · Claude Code · API
Setup IDE Rules↗
🧭

Turn your technical choice into a development budget

Compare 40 dev plans & simulate token costs vs $20/mo subscriptions

⚔️ Compare Model→💰 Compare plans & limits↗
Action HubReady to adopt this in production?

Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.

⚔️ Launch VS Compare→💳 View Plans↗
Up NextGroq Deploys NVIDIA Groq 3 LPX Alongside Vera Rubin NVL72 for Production Agent Inference
Read Now→