Subagent
Also known as child agent, delegated agent
In plain English
A subagent is an additional agent started by a main agent to handle a smaller piece of the same task, sometimes at the same time as other work.
Technical definition
A subagent is a delegated agent execution with its own conversation state and model requests, connected to a parent session through task and dependency relationships.
Engineering details
The main agent can launch one or more subagents and later consume their results. Their requests may overlap with the parent or with each other, creating branches in the session graph. Each branch can grow a separate context while sharing some initial instructions or repository state.
Why it matters
Subagents make agent traffic less sequential. A serving stack may receive bursts of long-context requests from one user task, and scheduler decisions affect how quickly branches finish. Aggregate throughput can rise while an individual branch waits longer for service.
How to read it in InferenceX
AgentX preserves subagent branches from the trace-derived workload and replays their dependencies. Delegation quality is outside its scope. The benchmark measures how the inference system serves the resulting parallel requests, shared prefixes, and completion timing.
Source material
See the concept in real benchmarks
Ultra-High Interactivity on NVIDIA GPUs? TileRT on InferenceX
Can TileRT software on NVIDIA GPUs compete with Cerebras, Groq LPU, and SambaNova? Batch size 1, disaggregated engine, high-throughput prefill engine, high-interactivity decode engine
Vera Rubin NVL72 vs GB200 NVL72? Inference TCO & Architecture Analysis
Rubin LUT Based Tensor Core, Feynman, Rack Scale, Perf Per MegaWatt, Perf Per Dollar, Software Improvements, Public Rubin Software, PyTorch, vLLM, OpenAI Triton