TL;DR
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Businesses are discussing and adopting open-weight AI models as the cost of using advanced AI services rises. Tinder says it has begun routing some non-technical users’ queries to open-weight models, while a Financial Times analysis cited a sixfold rise in corporate references to open-weight or open-source models in August and September compared with the same period in 2025.
The shift may give companies more control over costs and infrastructure, but the scale of adoption, realized savings and additional operating responsibilities remain unclear.
American businesses are exploring open-weight AI models as expenses for advanced AI services rise, with Tinder saying it now routes some queries from non-technical users to these models. The Financial Times reported Sept. 27 that corporate mentions of open-weight or open-source models rose sixfold in August and September compared with the same two-month period in 2025, citing AlphaSense data.
Open-weight models can be run on a company’s own hardware, potentially reducing reliance on paid access to advanced models offered by providers such as OpenAI and Anthropic. The Financial Times said the trend is most visible among technology companies, but reported that organizations in other sectors have also discussed using them. Those named include PNC Financial Services, logistics company C.H. Robinson and industrial company Siemens. The report does not describe the extent of each organization’s deployment.
Tinder CTO Vinay Kuruvila said the dating app had started routing some requests from non-technical users to open-weight models as it tried to manage rising AI costs. He said the company’s spending rate increased from $1 million annually in January to $10 million by July. Kuruvila said existing frontier models were capable enough for 90% of the tasks Tinder was trying to perform, and suggested that more capable open-weight models could reduce its need for those services. These figures and assessments are his account of Tinder’s experience.
The reported increase in corporate references comes from earnings calls and investor conferences, according to the Financial Times, which cited research platform AlphaSense. It is a measure of discussion, not a count of companies that have deployed the models or a measure of spending saved. The cited material does not specify how many companies made the references or provide an adoption rate.
Cost Control Brings New Choices
For businesses adding AI to more tasks, the choice of model can affect both technology spending and how much infrastructure a company must operate. Open-weight models may allow firms to run systems on their own hardware and select models for particular jobs. Tinder’s account illustrates one approach: sending some user requests to a different model rather than relying on frontier systems for every task.
The potential savings come with responsibilities. Companies may need to supply computing capacity, manage deployment and maintain safeguards and performance. A previous PYMNTS report framed the decision for middle-market finance leaders as a question of whether savings, flexibility and control justify taking on more responsibility for the underlying infrastructure. The current report does not quantify those costs or establish that open-weight models are cheaper in every use case.
The decision may matter beyond technology departments as AI expands into functions such as finance, procurement, treasury and compliance. If companies use these systems in operational workflows, cost and control decisions could affect how broadly they deploy AI. The available reporting describes interest and selected use, but does not show how far that broader shift has progressed.
Open and Closed Model Debate
The discussion reflects a wider debate over open versus closed AI systems. Open-weight models make their trained parameters available for users to run or adapt, while access to many proprietary models is provided through services controlled by their developers. The source material describes open-weight models as models companies can run on their own hardware; it does not claim that every model described as open source has the same terms or capabilities.
PYMNTS has previously described the divide as a financial and operational question for many companies, rather than simply an ideological dispute. Separately, Nvidia joined other technology companies in a safety coalition and called for shared open infrastructure for AI defense, including datasets, evaluation frameworks, attack simulators and red-teaming tools. That coalition concerns safety resources and is distinct from companies’ decisions about which models to run.
The current report points to a recent rise in business discussion: August and September 2026 compared with those same months in 2025. The comparison is based on mentions in earnings calls and investor conferences, as reported by the Financial Times from AlphaSense data. It does not establish when the trend began or whether discussion will translate into sustained adoption.
““invest in shared open infrastructure for AI defense — datasets, evaluation frameworks, attack simulators and red-teaming tools””
— Nvidia, in a statement cited by PYMNTS
Adoption and Savings Remain Unclear
The available reporting does not establish how many companies have moved from discussion to deployment, how much they have saved, or whether open-weight systems meet their needs across different tasks. The sixfold figure tracks references in business discussions over two-month periods; the underlying counts and a measure of actual adoption were not provided in the source material.
It is also unclear what additional infrastructure, staffing or safety costs companies incur when running models themselves, and how those expenses compare with paid access to proprietary systems. Tinder’s example covers some queries from non-technical users, not a reported replacement of its other AI services. Kuruvila’s view that open-weight models could eventually meet more of the company’s needs remains conditional.
Companies Weigh Deployment Trade-Offs
Companies considering open-weight AI will need to determine which tasks are suitable, what computing and operational resources are required, and whether the total cost compares favorably with paid model access. Further disclosures in earnings calls and investor conferences may show whether the recent increase in discussion is followed by broader deployment.
For Tinder, Kuruvila’s comments point to continued evaluation as model capabilities and costs change. The source material provides no timeline for expanding the routing of queries or a measure of resulting savings. The scale and financial effect of the wider corporate shift remain to be established.
Key Questions
Why are businesses considering open-weight AI models?
Companies are exploring them as a possible way to control rising AI expenses and gain more flexibility over where models run. The potential savings depend on each company’s infrastructure and operating costs.
Has Tinder replaced its other AI models?
No such replacement is reported. CTO Vinay Kuruvila said Tinder had begun routing some queries from non-technical users to open-weight models.
What does the sixfold increase measure?
According to the Financial Times, citing AlphaSense, references to open-weight or open-source models in earnings calls and investor conferences rose sixfold in August and September 2026 compared with the same months in 2025. The figure measures mentions, not deployments or savings.
Which companies outside technology have discussed these models?
The Financial Times named PNC Financial Services, C.H. Robinson and Siemens. The cited report does not detail the scale or status of each company’s plans.
Are open-weight models always cheaper?
The source material says they can cost less than paying for advanced model access, but does not establish that they are cheaper in every case. Companies also have to account for the infrastructure and other responsibilities involved in running them.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
