Vercel AI Gateway November Index: Decision models hit 26% of requests, Anthropic wins new teams 2:1

• Decision models account for 26% of all gateway requests in October 2026, representing 0.53% of total spend due to their low cost per request.
• Anthropic is the first-choice lab for nearly twice as many new teams as OpenAI, with both labs reaching nearly equal shares of paid teams (54% vs 55%).
• More than half of language-model requests go to models less than 60 days old, indicating a model half-life of approximately six weeks.
• Voice model adoption among paid teams increased by 161% since July, rising from 4% to 11%.

Decision Model Adoption

• Decision models classify, score, or pick choices rather than generating text, costing about 1/70th as much per request as language models.
• Jev from TypeSafe AI received nearly four times the total requests of Anthropic models, with teams spending $64 on Anthropic for every $1 spent on Jev.
• Adoption correlates with volume: 70% of teams processing over 10 billion tokens monthly used a decision model, compared to just over 20% of teams under 10 million tokens.

Model Churn and Vendor Competition

• A typical model peaks within one month of launch, and three in four teams using it at peak have moved on within four weeks.
• OpenAI's share of gateway spend increased by two-thirds since December, reaching nearly 15 cents per dollar, while Anthropic's share halved to about 32 cents.
• Opus 5.5 was the top model by spend in its first full month, taking about one in seven dollars across all teams.

Voice and Media Trends

• Speech-to-text and text-to-speech usage have tripled since July, with all-in-one voice models swapping places with assembled pipelines as the preferred approach.
• Image and video models had their largest months yet, with about one-third of the top 1% of teams by spend using an image model and one in six using a video model.