IBM Chief Says Open-Weight Models Now Make Sense for On-Premises AI
By Simon Sharwood · · TexaHub · Artificial Intelligence
IBM chief executive Arvind Krishna told investors that open-weight models cost 10 to 25 percent of frontier models' token cost while being about 90 percent as effective. IBM reported third-quarter revenue of $17.1 billion.
Earnings call remarks
Krishna said the benchmark gap between closed and open models has narrowed sharply since 2024, and that most enterprise tasks do not need the most expensive models. He said IBM will release software to ease on-premises AI deployments. [1]
IBM's quarterly revenue rose 5 percent to $17.1 billion, with software up 9 percent to $7.8 billion and mainframe revenue down 20 percent. [1]
Sources
[1] IBM CEO says open weight models are now irresistibly good and cheap on-prem — The Register