Open-weight models take 56% of token volume, Astra doubles Fable 5.1 spend
By Steven Van ·
Vercel's AI Gateway data shows Anthropic holding 64% of spend and per-token prices down 23.2% in August.
Anthropic kept 64% of all spend on Vercel's AI Gateway in August, even as open-weight models ran the majority of tokens for the first time: 56%, up from 7% in December.
- Price per token fell 23.2% in August, the third straight monthly drop. The median team paid 7.6% less, and the average token now costs less than half what it did five months ago.
- Fable 5, Anthropic's most capable model, lost two-thirds of its share of gateway spend in one month, while the cheaper Opus 5 tripled its share.
- Gemini 3 Flash has lost 95% of its share of gateway tokens since May, with more than three-quarters of that volume going to models from other labs.
- OpenAI's GPT-6 Astra, launched September 3, took 7.7% of all gateway spend in its first twelve days, more than twice the 3.7% that Fable 5.1 took after launching two days earlier at the same price.
Open-weight models accounted for just 14% of spend in August despite their 56% share of tokens. Full figures are in the AI Gateway Production Index.
Vercel
The platform for frontend developers — deploy, preview, and scale web apps and AI agents with zero config.
View Vercel →