Where LLM dollars actually go, by app category
Each OpenRouter app self-declares one or more categories (coding agent, personal agent, roleplay, productivity, etc.). We roll up every app's 30-day token volume and monthly cost into those buckets — an app with multiple categories is counted in each, so rollup totals exceed the global totals.
| Category | Apps | Tokens (30d) | Monthly cost | Blended $/M | % of spend |
|---|---|---|---|---|---|
| uncategorised | 27 | 66.84T | $0 | — | NaN% |
| cloud-agent | 3 | 64.5B | $0 | — | NaN% |
| game | 3 | 100.1B | $0 | — | NaN% |
| programming-app | 6 | 77.0B | $0 | — | NaN% |
| cli-agent | 6 | 50.0B | $0 | — | NaN% |
| video-gen | 2 | 15.4B | $0 | — | NaN% |
| roleplay | 1 | 6.2B | $0 | — | NaN% |
uncategorised
27 apps · 66.84T tokens · $0 monthly cost · — blended
cloud-agent
3 apps · 64.5B tokens · $0 monthly cost · — blended
game
3 apps · 100.1B tokens · $0 monthly cost · — blended
programming-app
6 apps · 77.0B tokens · $0 monthly cost · — blended
cli-agent
6 apps · 50.0B tokens · $0 monthly cost · — blended
video-gen
2 apps · 15.4B tokens · $0 monthly cost · — blended
roleplay
1 apps · 6.2B tokens · $0 monthly cost · — blended
Notes
- Double-counting: apps with multiple self-declared categories (e.g. "cli-agent" and "personal-agent") are summed into each. Rollup totals will exceed the global total on /agentic/openrouter-apps.
- Blended $/M is the category-weighted average — bigger spenders dominate the figure.
- Snapshot: 2026-08-24. Categories are whatever each app claimed to OpenRouter at the time of analysis.
Category misclassified?
OpenRouter's self-declared categories are imperfect. Spot an app in the wrong bucket? Tell us and we'll fix the rollup.