Discover how IBM Open-Weight AI Models could slash AI costs and why businesses are rethinking expensive AI solutions.

Businesses may no longer need to spend a fortune to get good results from AI. CEO Arvind Krishna questions why businesses should pay significantly more for advanced artificial intelligence when cheaper alternatives can deliver nearly the same results. During IBM’s third-quarter 2026 earnings call, Krishna said open-weight AI models now offer around 90% of the effectiveness of leading frontier models at just 10% to 25% of their token costs.

Open-Weight AI Models Challenge Expensive Alternatives

As Krishna pointed out, the gap between the open weight models and the frontier models is significantly reduced. The gap was higher in late 2024, where the frontier models were performing at a score of 70 on general benchmarks compared to the open models whose performance stood at 20 to 30.

The CEO believes this progress makes open-weight models a practical choice for many enterprise applications.

“That’s an easy question to ask of any enterprise,” stated Krishna, who posed the question of whether businesses would like a cheaper version of the product, which costs just one-tenth of the price but offers 90% of performance, or pay ten times as much for the highest version.

For routine work such as reading summaries, analyzing agreements, replying to client calls, and writing emails, IBM contends that the cheap versions can perform quite satisfactorily.

Why Are Enterprises Turning towards Private AI Infrastructure?

Besides cost considerations, there is the factor of control over confidential data, as AI becomes more and more embedded in business operations.

Krishna said an average large enterprise could begin running open-weight models with just two or three racks of infrastructure. This gives organizations another option beyond relying entirely on external AI services.

This strategy also encourages the recent trend towards sovereign AI, in which companies can exercise more control over their data, model execution, and operations logs. Implementing AI within a company’s own infrastructure can assist firms in complying with their particular security and governance requirements.

However, companies will have to take into account the cost of maintaining their own infrastructure, as well as their computing power and skills.

IBM Reports Stronger Revenue Growth in Q3 2026

IBM reported third-quarter revenue of $17.1 billion, while software revenue accelerated by 9%. Transaction processing also grew 10%, supported by stronger demand for IBM mainframe software and the completion of deals delayed from the previous quarter.

As Business Fortune observes, the company's overall revenue grew 6% at constant currency. Infrastructure revenue benefited from demand for distributed systems and storage, while Red Hat OpenShift annual recurring revenue increased by more than 30%.

Consulting remained a weaker area, with revenue growing 2%. Despite the fact that the share of consultancy income related to AI was at around 30%, growth within the sector has not been particularly significant. There have been some doubts regarding the valuation of the stock of IBM since early 2026.

FAQs

What are open-weight AI models?

Open-weight AI models make their trained parameters available for organizations to download, deploy, and, depending on their licences, modify. They can offer greater flexibility than models available only through external services.

Why does IBM believe open-weight models are more cost-effective?

IBM CEO Arvind Krishna said these models can deliver around 90% of frontier models' effectiveness at approximately 10% to 25% of their token costs, making them attractive for routine business tasks.

What is sovereign AI, and why does it matter?

Sovereign AI refers to AI systems operated under an organization's or jurisdiction's control, particularly regarding data, infrastructure, and governance. It matters when businesses need to protect sensitive information or meet regulatory requirements.

Can open-weight AI models replace frontier models completely?

Not necessarily. Open-weight models may handle routine tasks effectively, while frontier models can remain valuable for complex reasoning, demanding technical work, and applications requiring the highest available performance.

How could IBM benefit from the growth of open-weight AI?

IBM could benefit by helping businesses build and operate private AI infrastructure, integrate models into existing systems, and manage enterprise data securely. Demand will depend on whether customers see clear financial and operational benefits.

 

Sources: https://www.kucoin.com/news/flash/ibm-ceo-says-open-ai-models-offer-90-effectiveness-at-10-cost https://www.benzinga.com/markets/earnings/26/10/62517559/ibm-ceo-says-open-ai-models-are-90-as-good-as-frontier-models-why-pay-10x-more https://www.theregister.com/systems/2026/10/22/ibm-ceo-says-open-weight-models-are-now-irresistibly-good-and-cheap-on-prem/5304640