• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Do Chinese open-weight models really process most of the world’s AI tokens?

    A post argues that OpenRouter’s leaderboard is a misleading guide to global AI use, pointing to Ramp customers’ spending on OpenAI and Anthropic.

    Mary Clare McMahonMC
    1 Source, 23d ago, first seen 23d ago

    TLDR

    A post says Chinese open-weight models accounted for approximately 53% of OpenRouter’s public token volume as of September 13, 2026, but questions whether that reflects global use. Citing Ramp’s reported purchasing figures, it calculates that at least 77% of Ramp customers paying for AI subscribe to OpenAI or Anthropic. The exact share depends on the undisclosed overlap between those customer groups. It also says about 6% use platforms like OpenRouter—and 96% of that group also pay for OpenAI or Anthropic. From that adoption evidence, the post argues that closed models likely dominate AI deployment.

    Combined views

    7K

    1 Source, first seen 23d ago

    Combined views

    7K

    1 Source, first seen 23d ago

    34 likes
    34 likes
    2 comments
    19 saves
    7 reposts
    2 comments
    19 saves
    7 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    1 Source

    Mary Clare McMahon@mary_clare_mAre Chinese open-weight models really processing most of the world’s AI tokens? If you use OpenRouter’s AI Model Leaderboard as your reference, you might think so. As of September 13th, Chinese open-weight models accounted for approximately 53% of the platform’s public token volume.[1] But @KonstantinPilz and I think that data is misleading. Closed models dominate enterprise adoption. Of Ramp customers that pay for AI, at least 77% – and likely much more! – pay for an OpenAI or Anthropic subscription.[2] Only about 6% use platforms like OpenRouter, and the vast majority of those who do – 96%! – also pay for Anthropic or OpenAI. We expect that consumer AI use is even more closed-model dominated. So, it’s fair to guess that closed models dominate AI deployment by a wide margin. We dug into the evidence in a short post: [1] Of course, not all tokens are created equal. It’s not very meaningful to just compare token numbers without adjusting for model size and capabilities. Yet, it seems worth clarifying that even without adjustment, closed model tokens likely dominate. [2] The range is calculated using Ramp’s reported percentages: 55% of businesses purchased an AI product, while 42.4% purchased Anthropic and 39.5% purchased OpenAI. Because Anthropic users alone represent 42.4 ÷ 55 = 77.1% of AI-spending businesses, at least 77.1% used OpenAI or Anthropic. The exact figure depends on the undisclosed overlap between OpenAI and Anthropic customers.23d

    1 Source

    Mary Clare McMahon@mary_clare_mAre Chinese open-weight models really processing most of the world’s AI tokens? If you use OpenRouter’s AI Model Leaderboard as your reference, you might think so. As of September 13th, Chinese open-weight models accounted for approximately 53% of the platform’s public token volume.[1] But @KonstantinPilz and I think that data is misleading. Closed models dominate enterprise adoption. Of Ramp customers that pay for AI, at least 77% – and likely much more! – pay for an OpenAI or Anthropic subscription.[2] Only about 6% use platforms like OpenRouter, and the vast majority of those who do – 96%! – also pay for Anthropic or OpenAI. We expect that consumer AI use is even more closed-model dominated. So, it’s fair to guess that closed models dominate AI deployment by a wide margin. We dug into the evidence in a short post: [1] Of course, not all tokens are created equal. It’s not very meaningful to just compare token numbers without adjusting for model size and capabilities. Yet, it seems worth clarifying that even without adjustment, closed model tokens likely dominate. [2] The range is calculated using Ramp’s reported percentages: 55% of businesses purchased an AI product, while 42.4% purchased Anthropic and 39.5% purchased OpenAI. Because Anthropic users alone represent 42.4 ÷ 55 = 77.1% of AI-spending businesses, at least 77.1% used OpenAI or Anthropic. The exact figure depends on the undisclosed overlap between OpenAI and Anthropic customers.23d