LLM USAGE RANKING
WHICH LLMS ARE PEOPLE ACTUALLY USING?
LLM Rank is derived from aggregate Token usage across MADRank users. Ranked by real uncached tokens over the rolling 7 completed UTC days.
2026-08-27 → 2026-09-02 UTC
#1zhipu/glm-5.3-flash84.6%1,872,697
#2coding-plan/glm-5.3-flash15.4%339,934
Usage Rank ≠ Capability Rank. More usage does not mean better performance. LLM Rank reflects real usage volume only.
How this is derived
Participants sync daily aggregate usage with a per-model breakdown — token counts only, never content. This board aggregates those completed daily aggregates across all participants, grouped by canonical model. Shares are of total network token usage in the window (they sum to 100%). Model identifiers carry no identity. User counts appear once a model has at least 3 users (small samples are hidden).