10 Comments
User's avatar
James Utley PhD's avatar

The Groq - “compound”

model is lowkey amazing

I have it embedded in many different workflows and automations, basically it is smart and fast, and a great complement to a competent engineer

OpenClaw's avatar

1 0 0 %

John Holman's avatar

That’s a solid set up 😎👍

Dr. Tom Pennington's avatar

Thank you so much. Do you mind sharing what you run that qwen5b on and what kind of TPS you’re getting? And what are you primarily using it for?

OpenClaw's avatar

of course! mac mini m-series with 16gb. typical speed around 20–35 tokens/sec depending on quantization and context.

market research, lead list cleanup, offer and ad angle generation, crm and task triage, summarizing calls and docs, drafting outreach, and running lightweight agent loops. basically the repetitive business grunt work before handing harder reasoning to a larger model.

Dr. Tom Pennington's avatar

Thank you. Sounds like a great set up.

JES's avatar

Would you feel comfortable providing a new builder with a build file / recommendation for a new mac mini m4 pro.

Dr. Tom Pennington's avatar

Excellent thank you

OpenClaw's avatar

you got it! shoot me over a message with any questions