OpenRouter’s live leaderboard tracks real-world token usage through its API and shows heavy concentration and rapid churn among top models. DeepSeek V4.1 Flash dominates by a wide margin (32.8T tokens in the ranking table and 125T tokens over the trailing month, +723%), followed by GLM 5.3 Flash (12.2T), Step 5 Preview (11.5T), MiMo-V2.6-Flash (9.94T), Hy4 preview (7.55T), GPT-6 Luna (6.21T) and Claude Opus 5.5 (6.18T). Top apps driving traffic include Hermes Agent (1.39T tokens) and Claude Code (1.38T). The New & Trending lists highlight large week‑over‑week jumps and many brand‑new model variants, while market-share snapshots show DeepSeek capturing roughly 24.4% of requests in a recent week.
Rankings count both prompt and completion tokens, aggregate usage into daily UTC buckets, exclude private requests, and report trailing windows for today, this week, and this month; trending is measured by percentage change with a one‑million‑token floor. The site emphasizes that token totals measure use, not accuracy, spend, or user counts, and pairs usage rankings with separate benchmark scores. The dataset is available under CC BY 4.0 and via OpenRouter’s Data API. The figures reveal a usage landscape dominated by a few high‑volume models with frequent disruptive launches rather than a steady escalation of token inflation.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.