Meta's Internal "Claudeonomics" Leaderboard Revealed, 85,000 Employees Burned 600 Trillion Tokens in 30 Days

According to 1M AI News monitoring, Meta's intranet has seen an AI leaderboard named after Anthropic's flagship product Claude, called "Claudeonomics," built by employees using company data, aggregating token consumption data from over 85,000 individuals and listing the top 250. A copy of the leaderboard seen by The Information shows a total consumption of over 600 trillion tokens in the past 30 days, roughly equivalent to $9 billion based on a public average price of Claude Opus 4.6 (approximately $15 per million tokens), although Meta's actual model mix and pricing protocol are unknown. The top-ranked individual user has an average consumption of 2.81 trillion tokens, with costs potentially amounting to millions of dollars.

The leaderboard features gamified incentives, with progressive levels from Bronze to Jade, culminating in titles like "Token Legend" and "Session Immortal," as well as "Model Connoisseur" and "Cache Wizard." Some employees engage in rank chasing, causing AI agents to run consecutive hours of research tasks solely to inflate usage. Meta CEO Mark Zuckerberg and CTO Andrew Bosworth did not make it into the top 250.

Silicon Valley has recently embraced a "tokenmaxxing" trend, where token consumption is becoming a new measure of engineer productivity. In a tech conference back in February, Bosworth mentioned that the token spending of a top engineer equaled their salary, resulting in a productivity boost of up to 10x, calling it a "no-brainer deal to continue burning with no limit." NVIDIA CEO Jensen Huang stated last month that if an engineer earning a $500,000 annual salary doesn't consume at least $250,000 worth of tokens per year, he would be "deeply concerned."

Meta engineers currently use external models from Anthropic, OpenAI, and Google, as well as internal tools like MyClaw (Meta's version of OpenClaw) and the recently acquired Manus. Zuckerberg presented a "bold ask" to the engineering team earlier this year in an internal memo: to rewrite Meta's codebase to enable AI agents to directly read and modify the code.