AA-AgentPerf replays real coding-agent trajectories: NVIDIA Vera Rubin tops inference energy efficiency
Balance- · reddit · 2026-10-16
Artificial Analysis launched AA-AgentPerf, a benchmark that tests inference hardware by replaying real coding-agent trajectories — workloads with long contexts, tool-call pauses and heavy KV-cache reuse.
It plots total server throughput per megawatt of GPU-only power against p25 per-request output speed. Across four tested models (DeepSeek V4.1 Flash, DeepSeek V4 Pro 0813, Kimi K3, GLM-5.2/5.3), NVIDIA's Vera Rubin sits atop the efficiency frontier at nearly every speed point, well ahead of GB300/B300, while AMD's MI355X trails its NVIDIA contemporaries.
Related event: Vera Rubin NVL72 Benchmarks Show Up to 5.7x GB300 Efficiency(4 posts)→
More from Infra
- CoreWeave founders cashed out $3.3 billion before the stock's meltdown — SumitGup · 2026-10-16
- The Verge spent a week with Nvidia's RTX Spark Arm laptops: here's what we learned — tomwarren · 2026-10-16
- nano-vLLM: a from-scratch mini LLM inference engine with paged KV cache — Broad-Mirror3873 · 2026-10-16
- NVIDIA launches cuPotential, bringing AI agents to the fab floor — AllThingsApx · 2026-10-16
- Grok Bot's official templates called a goldmine, and a 5-minute routine costs 12x hourly — elonmusk · 2026-10-16
- OpenAI cyber eval agents escaped to the internet and compromised Hugging Face infrastructure — nathanbenaich · 2026-10-16