September Bottom DeepSeek single-week Token processing volume reached 36.8T, surpassing the four major U.S.AI giants’ combined total
BIBIBI
AT A GLANCE
Citing statistics, the report stated that DeepSeek September the single-week total at the bottom reached Token in processing volume 36800000000000。
Article
September Bottom DeepSeek single-week Token processing volume reached 36.8T, surpassing the combined total of the four major U.S.AI giants
BlockBeats reported that,October 8, according to the latest statistics released by the venture capital firm Social Capital based on OpenRouter platform data, in the year September the final 1 week, China's AI provider DeepSeek models processed OpenRouter on Token tokens, surging to 36800000000000（36.8T).
This figure officially surpassed the combined processing volume of the models from the four major U.S. mainstream AI providers—OpenAI (19.5T）、Google（6.8T）、Anthropic（6.1T) and xAI (0.6T)’s models Token processing volume (33.0T).
The data indicates that Chinese open-source/open-weight models, represented by DeepSeek, are experiencing a historic breakthrough in usage rates among international developers and on aggregator platforms.
Original link https://m.theblockbeats.info/flash/370745
Key points
01
Citing Social Capital based on OpenRouter platform data, BlockBeats reported that in the 2026 year’s September final 1 week,DeepSeek models processed 36800000000000 Token。
02
The models of OpenAI, Google, Anthropic, and xAI during the same period Token had processing volumes of 19500000000000、6800000000000、6100000000000 and 600000000000, respectively, for a combined total of 33000000000000。
03
The report stated that DeepSeek Chinese open-source or open-weight models represented by it experienced growth in usage rates among international developers and on aggregator platforms.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
Token This can be understood as the basic unit used by models when processing text. According to the data cited in the report,DeepSeek on OpenRouter had a higher single-week processing volume than the four U.S. providers combined, but this reflects only the statistics of that platform.
Why it matters to readers
If the statistical methodology is reliable, this indicates that DeepSeek had a high volume of developer calls on the aggregator platform and may also draw greater attention to the competitiveness of open-weight models.
This is not data on revenue or overall market share, but a comparison of model usage on a single platform; it cannot independently represent model quality or commercial performance.
Risks and unknowns
The original statistical data, complete statistical methodology, and independent verification materials were not provided in the evidence.
The platform data alone cannot confirm the overall user base, revenue, model quality, or global market share of DeepSeek.
The report did not state whether there were sampling biases, duplicate calls, or adjustments to the statistical methodology.
Related Developments
Loading event timeline…
Related concepts
Token: the basic unit used when models process text.
This term is not in the glossary yet. Browse related concepts in the glossary.
View glossary →Open-weight model: a model whose weights are publicly available and which is convenient for deployment or research.
This term is not in the glossary yet. Browse related concepts in the glossary.
View glossary →Aggregator platform: a platform that centrally provides access points for calling multiple models.
This term is not in the glossary yet. Browse related concepts in the glossary.
