OpenRouter data: Proprietary AI model query share plunges from 60% to 25%
Editorial responsibility
- No named human review is recorded for this page.
- Source reporting is collected, normalized, translated or condensed automatically when needed.
- Automatically published source-backed update
Data from OpenRouter indicates that the share of routed queries for proprietary AI models dropped from approximately 60% to 25% within months, as its routing system directs each request to the lowest-cost model expected to produce an adequate answer. This allows open-weight models to gain usage without customers manually selecting them. AT&T is shifting a rapidly growing share of its AI workloads from proprietary models to cheaper open models, reporting cost savings of up to 80%. The post suggests that if other large companies follow suit, OpenAI and Anthropic could lose much of the routine enterprise usage they currently charge for, potentially undermining AI lab revenue while sustaining cloud demand.
Source report
Data from OpenRouter indicates a significant shift in AI model usage, with the share of routed queries directed to proprietary models dropping from approximately 60% to 25% within a matter of months.
OpenRouter's routing system automatically directs each request to the lowest-cost model expected to deliver an adequate answer. This mechanism allows open-weight models to gain adoption without requiring customers to manually select them.
Case Study: AT&T
AT&T is increasingly shifting its AI workloads from proprietary models to cheaper open alternatives. The company reports cost savings of up to 80% as it allocates a rapidly growing share of its AI operations to open models.
Broader Implications
If other large enterprises follow AT&T's lead, AI labs such as OpenAI and Anthropic could lose a substantial portion of the routine enterprise usage they currently monetize.
This article was published by the Financial Times.
Source
rohanpaul_aiNeutral / independent