Gartner: Open-Weight Models Handle 90% of Agentic AI Work at Lower Cost
Gartner analysts report that open-weight AI models can perform 90% of routine steps in agentic workflows at one-twentieth the cost of frontier large language models. While current open models may require more retries, they still offer significant savings over proprietary systems. This shift challenges the economic model of for-profit AI providers who invest heavily in infrastructure for tasks that open models can handle efficiently.
Key points
- Gartner analysts Mark Petty and Raymond Paquet state that open-weight models can handle 90% of routine agentic workflow steps, such as classification and data validation.
- Open models cost approximately one-quarter of the cheapest closed model at equivalent performance levels and up to one-twentieth the cost of frontier models for full workflows.
- Current open models may be more verbose and require more retries, resulting in a cost saving of five to six times compared to frontier models rather than the best-case scenario.
- The report argues that for-profit model makers are chasing revenue from only the top ten percent of complex steps while ignoring the routine majority.
- Gartner rates the cost savings as a by-product of adopting open models, emphasizing that the primary value lies in the efficiency of handling routine operational executions.
The Economic Divide in Agentic AI
A new report from analyst firm Gartner challenges the prevailing economic assumptions in the artificial intelligence sector, suggesting that the heavy investment in frontier large language models (LLMs) may be misaligned with actual enterprise needs. According to the analysis, for-profit model makers are currently chasing revenue opportunities that constitute only ten percent of the work involved in agentic AI workflows. The remaining ninety percent consists of routine tasks that open-weight models can perform with parity to more expensive proprietary systems.
Routine Tasks Dominate Workflows
The report, titled Open-Weight Models Will Run 90% of Enterprise Agentic Workflow Steps, was authored by Gartner senior director Mark Petty and group vice president Raymond Paquet. They observe that typical operational workflows involve hundreds of routine executions. These tasks include classification, tool calling, passing data between systems, and validating or reconciling information. The analysts note that open-weight models are capable of handling hundreds of these routine steps for every single step that requires the advanced reasoning capabilities of a frontier model.
Cost Comparisons and Efficiency
Gartner provides specific cost metrics to illustrate the disparity between open and closed models. When compared to the cheapest closed model at the same performance score, an open model costs approximately one-quarter as much. In scenarios where a frontier model is run on every step of a workflow, the cost saving for using an open model can be as high as one-twentieth. These figures represent the best-case scenarios for open-weight adoption in enterprise settings.
Real-World Performance Constraints
Despite the theoretical cost advantages, the analysts acknowledge practical limitations in current open models. They admit that open models can be more verbose and may retry operations more often than their closed counterparts. Consequently, the actual saving per task against a frontier model is closer to five or six times rather than the twenty-fold saving seen in ideal conditions. Gartner rates these cost savings as a welcome by-product of moving to open models rather than the primary end goal of the technology.
Strategic Implications for SaaS and AI Providers
The report highlights that years of workflow history represent a significant portion of what acquirers pay for in a Software-as-a-Service (SaaS) company. The implication is that the value lies in the execution of routine, high-volume tasks rather than just the ability to perform complex reasoning. By shifting the burden of routine steps to cheaper open models, enterprises can optimize their spending while maintaining operational efficiency.
The Future of Model Deployment
Gartner’s analysis suggests a fundamental shift in how AI infrastructure should be deployed. Instead of relying on expensive frontier models for all tasks, organizations are encouraged to use a hybrid approach. Open-weight models should handle the bulk of routine executions, while frontier models are reserved for the complex, high-value steps that truly require advanced reasoning. This strategy aims to reduce costs without compromising the quality of the final output.
Key takeaways
Gartner analysts Mark Petty and Raymond Paquet argue that open-weight models can handle 90% of routine agentic workflow steps.
Open models cost approximately one-quarter of the cheapest closed model and up to one-twentieth of frontier models in best-case scenarios.
Current open models may require more retries, resulting in a five to six times cost saving compared to frontier models.
For-profit AI providers are focused on the top ten percent of complex steps while ignoring the routine majority.
Gartner views cost savings as a by-product of open model adoption, emphasizing efficiency in handling routine operational executions.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.
