Vercel Data Shows AI Models Have Six-Week Half-Life in Production

Vercel's November 2026 AI Gateway Production Index reports that AI models in production have an effective half-life of about six weeks, with more than half of October's language-model requests going to models less than 60 days old, up from less than three in ten before September. Vercel, whose AI Gateway routes more than 100 trillion tokens a month, found that a typical model peaks within a month of launch and that four weeks later three in four teams using it at peak have moved on, while the median model family ships a new version every 61 days. The index also shows decision models reached 26% of all gateway requests by October on just 0.53% of spend, voice use rose to 11% of paid teams from 4% in July, and Anthropic closed to 54 paid teams per 100 against OpenAI's 55, up from 48 versus 70 last December.

Vercel Data Shows AI Models Have Six-Week Half-Life in Production Deployment platform Vercel reports that AI models in production are being replaced so quickly they have an effective half-life of about six weeks, according to its November 2026 AI Gateway Production Index. The finding suggests enterprise AI adoption is far more fluid than lab benchmarks imply, with teams switching models almost as fast as labs release them. Models peak fast, fade faster Vercel's AI Gateway routes more than 100 trillion tokens a month between production applications and model providers, giving it a view of actual usage across language, decision, voice, image, and video models. The November index shows that more than half of October's language-model requests went to models less than 60 days old, up from less than three in ten before September. A typical model peaks within a month of launch; four weeks later, three in four of the teams that were using it at its peak have moved on—whether or not a successor has shipped in the same family. Most of those teams remain on the gateway, just using other models. "AI models now have a half-life of about six weeks," Vercel writes. "That means that, to the teams using them, labs are only as good as their latest model and have to re-earn their business with every new one." The median model family ships a new version every 61 days. Teams adopt new models within days of release: in September, about one in six language-model requests went to models less than two weeks old. Among teams running at least 10 million tokens a month, roughly half changed their most-used model in October. Four in five of those switches went to a more recent release, and six in ten moved to a different lab. Older models persist but shrink to a sliver of traffic. Seven in ten high-volume teams still used a model at least ten months old in October, but for the typical team it accounted for less than 3% of tokens. Decision models take a quarter of requests Production AI is splitting into two jobs: generation and judgment. Decision models classify, score, or pick between choices instead of generating text, and cost about a seventieth as much per request as a language model. They arrived on AI Gateway in mid-September; by October they accounted for 26% of all gateway requests while representing just 0.53% of spend. Nearly a third of paid teams used a decision model, and about 11% used one without sending a single request to a text-generating model. Adoption skews toward volume: just over one in five teams under 10 million tokens used a decision model, rising to nearly half at 100 million–1 billion tokens, and seven in ten above 10 billion tokens. Almost all of that request volume went to Jev from TypeSafe AI, which alone received nearly four times Anthropic's total requests. Voice adoption accelerates About 11% of paid teams used a voice model in October, up from 4% in July—a 161% increase. Among teams that paid in both months, voice usage rose from 5% to 14%. The share of teams sending at least 100 voice requests a month grew fivefold, and six in ten of September's voice teams returned in October, up from fewer than half in July. Speech-to-text is the most common type; its use has nearly tripled since July, and text-to-speech has more than tripled. Teams are also shifting from all-in-one voice models toward assembling separate speech-to-text, language, and text-to-speech components. Anthropic wins new teams; OpenAI grows spend share Last December, 70 of every 100 paid teams used an OpenAI model and 48 used Anthropic. By October the two were nearly level at 55 and 54 per 100. Anthropic's gain came largely from new teams: in October it was the first-choice lab for nearly twice as many of them as OpenAI, and has held that lead in nine of the last 11 months. Among the top 1% of teams by spend, more than nine in ten use both labs, so competition has shifted to share of wallet. OpenAI now takes nearly 15 cents of every dollar spent on the gateway, up from about 9 cents in December, while Anthropic's share has halved to about 32 cents over the same period. Image and video spending hit records Image and video both had their biggest month yet on AI Gateway. About a third of the top 1% of teams by spend now use an image model, up from a quarter in July; one in six use a video model, up from one in ten. OpenAI's GPT-Image-2.5 Sunburst led image spend in its first full month; ByteDance's Seedance 2.5 led video spend for a third month running. What it means The six-week half-life implies that model providers cannot rely on lock-in from integration effort or institutional inertia. Teams are switching frequently, often across labs, and keeping older models only as fallbacks. For enterprises, the operational challenge is not choosing a model but building pipelines that can swap models continuously without breaking downstream logic. For labs, the pressure is to ship improvements faster than the half-life—or risk losing the next cycle's traffic entirely. Sources 1. Decision models take 26% of requests; Anthropic now wins ... https://vercel.com/blog/ai-gateway-production-index-november-2026 vercel.com 2. Vercel estimates AI models have an effective half-life of ... https://mezha.net/eng/news/mezha-e34954ec6a277fb5427844d8 vercel estimates ai mezha.net 3. Gemini 4 Carbon is now available https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available blog.google 4. Introducing Claude Fable 5.2 and Claude Mythos 5.2 https://www.anthropic.com/claude-fable-and-mythos-5-2 anthropic.com 5. Next.js 16.4.5, AI SDK 7.0.165, TanStack AI 1.2 https://dev.to/sushilmagare10/november-2-8-2026-nextjs-1645-ai-sdk-70165-tanstack-ai-12-6b3 dev.to