The Pulse
Decision models took 26% of October AI requests, Vercel says
Decision models received 26% of AI Gateway requests in October, but accounted for just 0.53% of estimated spending, Vercel says. Nearly a third of paid teams used one, and adoption was higher among teams processing more tokens.
AI.info Team · Written with AI · fact-checked ·
Decision models gain request share
Vercel says decision models received 26% of requests routed through its AI Gateway in October, while accounting for 0.53% of total spend. The figures come from the company’s November 2026 Production Index, which reports anonymized, aggregate traffic routed through the gateway through October.
AI Gateway routes more than 100 trillion tokens each month between production applications and AI labs. Decision models are tools that classify, score or choose among options rather than generate text; they cost about one-seventieth as much per request as a language model, the company says.
The models became available on AI Gateway in mid-September, and 15 were available from eight vendors by the time of the report. Nearly a third of paid teams used a decision model in October; about 11% used one without sending any request to a text-generating model, according to Vercel.
Adoption rises with token volume
Adoption increased with team token volume in October, according to Vercel’s figures. Just over one in five teams processing fewer than 10 million tokens used a decision model, compared with nearly half of teams processing 100 million to 1 billion tokens and seven in ten processing more than 10 billion.
The largest teams use more AI models overall, the company says: the top 1% by spend used a median of 29 different models, while the bottom tenth used one. Teams have more to gain from handling simple decisions with a faster, lower-cost model as they send more requests, Vercel says.
Jev takes most decision-model traffic
Almost all decision-model requests went to Jev from TypeSafe AI, which Vercel says received nearly four times Anthropic’s total requests. For every $1 teams spent on Jev, they spent $64 on Anthropic models, according to the report.
Decision models’ share of requests was much larger than their share of spending. Their lower cost per token and short answers account for the difference, Vercel says: the models received about one in four gateway requests in October but made up just 0.53% of total spend.
Other shifts in the October index
About 11% of paid teams used a voice model in October, up from 4% in July, Vercel reports. Anthropic and OpenAI reached about the same share of paid teams by October, while new teams chose Anthropic nearly twice as often as OpenAI, the company says.
More than half of October’s language-model requests went to models less than 60 days old, according to the index; before September, that share had never passed three in ten. Vercel estimates spending using published list prices and says actual bills may differ.