Raw SignalSep 25, 20261 min read1 source

Why AI Assistants Are Slow—and How to Make Them Faster

In one minute

Cerebras published recorded benchmarks of AI assistants completing the same dinner-reservation request.

Its examples report Meta Muse at 4 minutes 36 seconds, Claude Cowork at 6 minutes 25 seconds with 57 tool calls, and Grok Bot at 7 minutes 40 seconds. Cerebras’ own Qwen 3.8 27B assistant, using the Pi harness, recorded a 22-second median across two successful attempts.

The company attributes the optimized path to parallel availability checks, a saved website-navigation procedure, and faster inference. It reports that its browser/API segment took 6.8 seconds versus 4 minutes 31 seconds in an earlier Grok trace, while tool calls fell by more than 80 percent. Cerebras cautions that this bundles model, harness, and execution-path changes rather than isolating any one intervention.

Evidence / sources

  1. 1cerebras.ai/blog/the-rise-of-slow-personal-assistants
All Raw Signals

Next reads

Read next

Open the graph
  1. 01archiveRaw SignalsOpen every source-linked raw signal.
  2. 02source ledgerSource LedgerInspect the public source evidence graph.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract