- Chinese AI models have surpassed US counterparts in token usage since early 2024, OpenRouter data shows
- Cheaper energy and efficient models help Chinese AI labs charge less per token than US firms
- DeepSeek V4 Flash leads token consumption with 4.63 trillion tokens, followed by MiniMax M3 and Xiaomi’s model
Chinese artificial intelligence models have far outpaced their US counterparts when it comes to token usage, OpenRouter data has revealed. Since the start of the year, Chinese AI models have overtaken US ones in token consumption, the aggregation platform found.
Cheaper energy and more efficient models have allowed China's AI labs to charge less than leading US companies for tokens, giving the country a new edge on the AI battleground, as per Financial Times. OpenRouter stated that only one US company was in the top five when it came to token usage- Anthropic.
The most popular model was DeepSeek V4 Flash with 4.63 trillion tokens, followed by MiniMax M3 with 4.13 trillion tokens and Xiaomi's MiMo-V2.5 with 3.8 trillion tokens. Tencent's Hy3, Anthropic's Claude Opus 4.7, DeepSeek V4 Pro, Owl Alpha, Claude Sonnet 4.6 and Claude Opus 4 round up the top 10.
Google's Gemini 3 and OpenAI's GPT 5.5 are in the 12th and 13th position in terms of global token usage, as per OpenRouter. In terms of weekly token usage, OpenRouter reported that 47.1 trillion tokens were used in the week ending June 15 compared to 44.6 trillion for the previous week. DeepSeek made up the biggest portion of usage in terms of specific AI models.
The difference between the US and Chinese models comes as several US AI firms are shifting to token-based billing, meaning they track the units of data processed by models. The move from flat subscription to another type of billing has exposed corporate clients more directly to the cost of every AI tool they use.
ALSO READ: Trump No Longer Views Anthropic As National Security Threat: Here's Why
Companies Put Caps On AI Usage
Now, firms which encouraged their workers to use AI tools are starting to rein in those mandates, as the cost of deploying artificial intelligence at scale has started testing corporate budgets, as per FT. Amazon, Cisco, Uber, Walmart and Meta have introduced caps or pushed employees to cheaper models to keep AI spending costs under control.
Goldman Sachs predicted that by 2030 the use of AI agents would lead to a 24-fold increase in token consumption, exacerbating the shortage of chips over the next 12 to 18 months.
As token usage and AI spending by companies increases, efforts to curb costs could impact the growth of startups like Anthropic and OpenAI, which plan to go public later this year at close to trillion-dollar valuations.
While some firms are shifting to open-source and cheaper models, many are now reconsidering the impact of AI on the productivity of employees and their own bottom line.
ALSO READ: Amazon Turns 'Water Positive' In India Amid Big Tech's Global Scrutiny Over AI Data Centres
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
