Comment by spunker540

The gpu inference happens on anthropics side, but the output of the gpu inference is often many local tool calls. The agent is locally reading files, writing files, compiling programs, downloading artifacts, running tests, invoking CLIs, monitoring logs, inspecting web sites, grepping etc, and can be doing all of that at once, in parallel via subagents, much faster than a single human ever could.