OpenRouter基于真实API调用数据发布了大语言模型排名榜单，按照通过其平台处理的token数量对模型进行排序1。该榜单覆盖日、周、月等多个时间维度，为行业提供了量化的模型使用情况洞察1。
截至2026年10月12日开始的一周，DeepSeek以24.4%的请求份额在排名中领先1。OpenRouter强调，这份榜单统计的是通过其API处理的prompt和completion tokens的总量，而非衡量模型的准确性、推理能力或基准性能1。榜单的趋势对比采用相邻七天的方法，仅纳入当前窗口内至少处理了100万token的模型1。
OpenRouter has published a ranking of large language models based on actual usage data processed through its API, ordered by token volume rather than performance metrics 1. The rankings reveal that DeepSeek leads the market, commanding 24.4% of request volume during the week beginning October 12, 2026 1.
The platform tracks both prompt and completion tokens across multiple time horizons—daily, weekly, and monthly—to capture usage trends 1. However, the rankings explicitly do not measure model accuracy, reasoning capability, or benchmark performance 1. OpenRouter calculates trend comparisons using a seven-day rolling window methodology and includes only models that have processed at least one million tokens within the current time period 1.
The title "Token-maxxing is dead" suggests that the previous growth trajectory of token consumption is stabilizing, reflecting a shift in how the market values and utilizes language models 1.
评论
还没有评论，欢迎留下第一条。