The AI Token Leaderboards, Compared
A year ago there was one public AI token leaderboard. After tokenmaxxing went mainstream in 2026, there are at least nine. We build one of them, so read this with that in mind — but we'd rather map the landscape honestly than pretend the others don't exist. Here's what each board does, who it's for, and how to pick.
The field at a glance
| Board | Scale | Tools covered | Identity |
|---|---|---|---|
| Viberank | 1,100+ devs · $10M+ tracked | Everything ccusage tracks (15 sources) | GitHub OAuth + verified badges |
| tokenmaxxing.sh | Smaller board, whale-heavy top 10 | ccusage sources | Account sign-in |
| TokenRank | Growing | Codex, Claude Code, Gemini, Qwen, Cursor, Copilot + more | X (Twitter) identity |
| Straude | Growing | Claude Code + Codex | Account sign-in |
| ccclub | Private friend groups | Claude Code, Codex, OpenCode, Amp, pi-agent | Invite codes |
| ccrank | Self-hosted instances | Claude Code (via ccusage) | Google OAuth, invite-only |
| CCWarriors | Growing | ccusage-readable agents | Account sign-in |
| DevBurn | Early | Claude Code (Codex/Cursor planned) | Account sign-in |
| Token Tracker | Dashboard-first | 27 tools | Opt-in display names |
The one-line difference for each: Viberank — the original and largest — tiers, profiles, readme badges, autosubmit, calculator, monthly data reports; tokenmaxxing.sh — clean spend/token windows (7d/30d/all), multi-device profiles; TokenRank — time-window boards, english/chinese, rank-challenge links; Straude — strava-style streaks and pace framing; ccclub — leaderboard among friends, not the public internet; ccrank — open-source team leaderboard you deploy yourself (cloudflare workers); CCWarriors — live-updating board with per-tool filters; DevBurn — points formula rewarding consistency, not just spend; Token Tracker — local-first dashboard with an optional public board.
How to pick
- You want the biggest cohort to rank against — the whole point of a leaderboard is the other people on it. Viberank has 1,100+ developers and $10M+ tracked; see the live stats.
- You want a private league with friends or your team — ccclub (invite groups) or self-hosted ccrank.
- You want streak/fitness framing over raw spend — Straude or DevBurn's points formula. (Viberank tracks streaks and active days on profiles too.)
- You want X-native identity — TokenRank ties ranks to X handles; Viberank ties them to GitHub, which is also what powers its verified badges.
What actually separates boards
Every board reads the same local data (usually via ccusage), so the real differences are cohort size (who you're ranked against), validation (whether absurd submissions get caught — Viberank sanity-checks token math, cost ratios, and dates server-side), identity (anonymous numbers vs verified GitHub accounts), and what exists around the rank — profiles with daily charts, README badges, tool comparisons, a plan calculator, and monthly data reports that press can cite.
Try the original
npx viberank-cliOne command, all your agents, only totals leave your machine. Then check whether your number is normal.
Leaderboard FAQ
What is an AI token leaderboard?
A public ranking of developers by their AI coding usage — tokens consumed and API-equivalent cost — usually measured by ccusage from local agent logs (Claude Code, Codex, Gemini CLI, etc.) and submitted as aggregate totals. Code and prompts stay local on every board listed here.
Which AI token leaderboard is the biggest?
Viberank, with 1,100+ developers and over $10M in tracked API-equivalent spend since 2025 — it's also the board that's been cited in press coverage of AI coding costs. Most alternatives launched in 2026 and are still building their cohorts.
Can I be on more than one leaderboard?
Yes — they all read the same local ccusage data, so submitting to several is just running several CLIs. The main cost is that each board is another place your usage totals live.
Do these leaderboards see my code or prompts?
No. Every board here submits aggregate rows only — dates, token counts, costs, model and tool names. Verify it yourself: most CLIs (including npx viberank-cli) are open source or offer a dry-run that prints the exact payload.