Coinbase's CEO wants his engineers to keep tokenmaxxing — while keeping AI costs down.
In an X post on Friday, CEO Brian Armstrong outlined five ways in which the crypto exchange is keeping AI costs low.
The first of his five strategies was selecting better default LLMs — the models most engineers use by default when submitting prompts. He said Coinbase was experimenting with Chinese LLMs as defaults, which are significantly cheaper than models from frontier American AI labs like Anthropic and OpenAI.
"We're experimenting with defaulting to open weight models like GLM 5.2 and Kimi 2.7 through our LLM gateway, while still encouraging engineers to choose the right model for the task," Armstrong wrote.
GLM 5.2 and Kimi 2.7 are models developed by the Chinese AI labs Z.ai and Moonshot AI, respectively.
His second strategy, one he had spoken about earlier in June, is routing prompts to their most appropriate models based on their difficulty levels.
"For instance, you may want a frontier model for planning, but not for execution where they can be overkill," he wrote. "Ultimately, humans shouldn't be choosing models - AI can automate this task."
The third tip was to use better caching, a technique that reduces inference costs.
The fourth was to keep context lean, meaning starting new sessions when switching between tasks.
And his final strategy is to improve visibility into AI spending across the company. This means all his engineers can use as many tokens as they want, but they can see their usage. Coinbase will expect "more impact" from employees who spend more on AI.
Armstrong attached a graph at the end of his post tracking token usage and AI spend at the company over time, though the exact timeline was not specified. The graph shows that token usage has recently reached one of the highest levels in the company's history, while AI spending has fallen significantly, to nearly half its peak level.
"The goal isn't to suppress usage. It's to build the infrastructure that makes exponential growth sustainable," he wrote.
Armstrong's post comes less than two months after Coinbase laid off 14% of its staff, partly due to AI changing how people work.
"Over the past year, I've watched engineers use AI to ship in days what used to take a team weeks," Armstrong said in a post in May, adding that, "the pace of what's possible with a small, focused team has changed dramatically."
His strategy to reduce AI costs falls in line with the industry, which has moved on from the short-lived tokenmaxxing trend in favor of imposing usage caps on employees to curb rampant token consumption.
