Decision Models Reach 26% of Vercel AI Gateway Requests
Decision models handled 26.1% of requests through Vercel AI Gateway in October, but accounted for just 0.53% of estimated spending, according to the company's November 9 Production Index. The figures describe traffic on Vercel's service, not market-wide AI adoption.
These models classify inputs, assign scores, or select among options instead of generating extended text. Vercel says the category arrived on its gateway in mid-September and now includes 15 models from eight vendors. Its request share rose from 12.8% in September to 26.1% the following month.
Nearly a third of paid gateway teams used a decision model in October. Adoption reached roughly seven in ten among teams processing more than 10 billion tokens monthly. Almost all decision-model request volume went to TypeSafe AI's Jev, making the category's growth heavily concentrated in one offering.
Request volume is not spending share
The gap between usage and spending matters for developers choosing how to divide application workloads. A high number of classification or routing calls can coexist with a small share of the budget, while text-generation models still account for substantially more spending. The report does not establish that decision models can replace general-purpose language models across tasks.
Vercel estimates spending using published list prices, so actual bills may differ. It defines a paid team as one with at least one paid request that month. Its methodology page also notes that a team is an account workspace, and one company can operate several.
Together, those limits make the report a useful snapshot of gateway workloads, rather than a count of companies or an independent industry census.