Post by DeepInfra

NVIDIA just published AgentPerf. The first open benchmark designed specifically for agentic AI workloads. DeepInfra is one of the inference providers featured in the results. Why this benchmark matters: most AI benchmarks measure a single model call. But agents don't work that way. They make sequential LLM calls, run tool calls in between, and often handle many concurrent sessions at once. AgentPerf measures all of that together. We're already seeing NVIDIA Blackwell perform in production agentic workloads. Pam.ai, an AI workforce platform for car dealerships that books service appointments, handles calls, and runs outbound sales campaigns, runs their agentic workflows on DeepInfra using gpt-oss-120b on NVIDIA Blackwell. If you're building agents, this benchmark is worth reading. Learn more: https://lnkd.in/g3szEjb3