Open-source AI models now run fast enough for real use
Original: RT @ClementDelangue: I remember when people were saying "It's useless to open-source big models because nobody will be able to run them fas…
Source: x.com ↗
Who: Posted by Bill Gurley (prominent Silicon Valley venture capitalist, known for Benchmark fund investments), retweeting Clément Delangue (CEO of ).
What's new: The tweet is a short victory-lap observation. Delangue recalls a time when skeptics argued that AI models were pointless because ordinary people and companies lacked the computing power to run them at useful speeds. That argument has now collapsed.
Why it matters: The skeptics were wrong on two counts. First, the hardware available to businesses and researchers has grown dramatically. Second, and more importantly, the software tools for running large models efficiently have improved just as fast — meaning the same model that once required a warehouse of computers can now run on far more modest hardware. This is the problem, and it has largely been solved by the open-source community itself.
Caveats: The tweet is a fragment — the full text is cut off — so the specific models or tools Delangue had in mind are not named. The claim is broadly true but glosses over the fact that the very largest remain out of reach for most organizations. Open-source has narrowed the gap, but has not eliminated it.