Decision models take 26% of requests; Anthropic now wins new teams 2:1
Source
vercel.com
Author
Amelia Charles
Date
Why it matters
Production data shows decision models at 26% of requests but about half a cent per dollar, and that most teams move off a model within a month of its peak. Useful for routing and model-selection decisions.
Key takeaways · AI-distilled
Vercel says decision models, which classify, score or pick between options instead of generating text, cost about a seventieth as much per request as a LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition →. About 11% of paid teams in October used one without sending any request to a text-generating model.
Almost all decision-model traffic went to TypeSafe AI's Jev, which received nearly four times Anthropic's total requests on the gateway. Spend ran the other way: teams spent $64 on Anthropic models for every $1 on Jev.
Voice builders are moving from all-in-one speech models to assembled pipelines. In July about a quarter of voice teams used an all-in-one model and one in six chained speech-to-text, a language model and text-to-speech; by October those shares had swapped.
Anthropic still leads gateway spend at about 32 cents per dollar, half its December share, while OpenAI rose from about 9 to nearly 15 cents. Among the top 1% of teams by spend, more than nine in ten already use both labs.
In October about half of teams running 10M+ tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → a month changed their most-used model; four in five moved to a newer release and six in ten to a different lab. Seven in ten still ran a model at least ten months old, but for under 3% of tokens.
Terms in this piece · Glossary
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
