Back to feed
News Story
APriority77
DeepTech深科技
1 sources

Open-Source AI Captures Majority of Tokens, DeepSeek Takes One-Fifth

According to Vercel AI Gateway data, the token share of open-weight models rose from 28.4% to 62% in two months, while closed-source models declined. DeepSeek's token share grew from under 1% to 22.6%, indicating that low-cost models are crossing the quality threshold for production use. Enterprises are increasingly selecting models based on task requirements, shifting the AI market competition toward cost-effectiveness.

SynthePulse Insight · AI deep readingMembers

Open-Weight Models Take the Token Majority: AI Competition Shifts from 'Best' to 'Best Value'

Version 1 · 1 source

Vercel AI Gateway data shows open-weight models' token share surged from 28.4% to 62% in two months, yet they account for less than 4% of spending. The focus of AI competition is shifting from model performance to matching tasks with price.

  • On August 22, open-weight models held a 62% token share on Vercel AI Gateway, up from 28.4% two months earlier; closed-source models fell from 71.6% to 38%.
  • In June 2026, open-weight models handled 29% of tokens but accounted for less than 4% of spending, with an average token price about one-tenth of the platform's overall average.
  • Agents amplify demand for low-cost tokens: In April, only 22.2% of requests included tool calls, yet they consumed 58.9% of tokens.
Open section navigationThe Dramatic Reversal in Token Share

The Dramatic Reversal in Token Share

On August 22, open-weight models' token share on Vercel AI Gateway, an AI model aggregation and routing platform, rose to 62%. Two months earlier, that figure was only 28.4%. During the same period, closed-source models' share fell from 71.6% to 38%. This data comes from a production platform covering hundreds of models, and while it cannot be directly equated with the entire AI market, it reveals a significant shift: enterprises are no longer assigning all tasks to the strongest, most expensive models.

This reversal is not due to a single event but rather a rapid adjustment in developers' model choices. On Vercel AI Gateway, developers are migrating high-frequency, low-risk tasks to open-weight models, while closed-source models are reserved for more complex scenarios. This divergence is especially evident in call volume and spending.

Free for now

Read the full analysis

4 more sections of analysis, plus the full takeaway

Loading

Credibility boundary

The data in this article primarily comes from public data from Vercel AI Gateway and Ramp, as reported by DeepTech深科技. Vercel AI Gateway's data represents only its platform and cannot be directly equated with the entire AI market. Ramp's data is from a third-party source, and its statistical methodology is not detailed. All data are as stated by the sources and have not been independently verified.

Sources for this version

  1. 开源AI拿下多数Token，DeepSeek占1/5DeepTech深科技

Primary report

DeepTech深科技

Primary source