DeepSeek достиг 22,6% токенов AI Gateway, open-weight модели заняли 29% — отчёт Vercel

Vercel published its AI Gateway Production Index report for June 2026 in July 2026: open-weight models grew to 29% of all tokens while accounting for less than 4% of spending, DeepSeek climbed to third place with 22.6% of volume, and Anthropic captured 61% of all platform spend.

Why open-weight models surged

Open-weight models drove 29% of all AI Gateway tokens in June 2026 — up from 11% in April, meaning the share nearly tripled in two months. Yet they accounted for less than 4% of the money spent: almost a third of traffic for one twenty-fifth of the budget. According to Vercel, the main contributor was DeepSeek — now the third-largest source of tokens on the platform by volume, with a 22.6% share, trailing Google by less than two points.

*Report for July 2026, data collected in June 2026

*Token volume grew 29% month-over-month, spending grew 27%

*Open-weight: 29% of tokens on <4% of spending (was 11% in April)

*DeepSeek — 22.6% of volume, third place, Google's share fell to 24%

*Roughly every eighth enterprise customer runs an open-weight model in production

A newcomer is accelerating the trend. Z.ai released GLM 5.2 — an open-weight model under the MIT license for long agentic tasks, priced at roughly five times cheaper than Opus 4.8. Since its API launched on June 16, 2026, through the end of the month, its daily token volume grew by roughly 50x, and the model itself climbed to 11th place on the platform by tokens in the final week.

Who dominates spending

Anthropic captured 61% of all AI Gateway spending on 32% of tokens and held 72% or more of spending in every high-stakes use case. This illustrates the routing discipline the report highlights: bulk work goes to cheap models, while critical work stays on the frontier.

"High-volume work moves to low-cost models, while high-stakes work stays on the frontier," the

Vercel AI Gateway report states.

The split is also visible across modalities. OpenAI's GPT Image generated 53% of images, ahead of Google's Nano Banana (39%). In video, Chinese labs captured two-thirds of spending, and xAI's Grok Imagine produced 42% of all clips — more than any other model.

What's happening with price per token

The average price per token in June 2026 stayed flat — after rising nearly 20% in May. Volume and spending grew at nearly the same rate (29% and 27%), so the average price per token didn't shift.

Two opposing factors were at play. Cheap open-weight traffic pulled the average price down, but at the same time frontier models got about 12% more expensive per token. They offset each other. One detail of the month stands out: Anthropic's Claude Fable 5 launched on June 9, 2026, and within four days reached 22% of Opus 4.8's request volume, but a US export control directive suspended access to it for the rest of the month.

What it means

AI investment keeps growing fast, but companies are spending smarter: cheap open-weight models handle volume, while expensive frontier models handle critical tasks. An open-weight lab could soon become the second-largest source of tokens on AI Gateway by volume.

FAQ

What are open-weight models?

These are models with open weights that a company can run on its own infrastructure — for example, DeepSeek or Z.ai's GLM 5.2. According to Vercel, in June 2026 they delivered 29% of the platform's tokens for just 4% of spending.

Which model leads by token volume?

By volume, Anthropic and Google hold first and second place, while DeepSeek ranks third with a 22.6% share — less than two points behind Google, whose share fell to 24%.

What is GLM 5.2?

GLM 5.2 is an open-weight model from Z.ai under the MIT license for long agentic tasks, roughly five times cheaper than Opus 4.8. After its API launched on June 16, 2026, its daily token volume grew roughly 50x within two weeks.