Data Shows Grok Leading Over ChatGPT in Agentic Trading

AI-driven payment activity is accelerating with the AI models now powering agentic trading.

Artemis data shows agentic-payment activity through protocols such as x402 and the Machine Payments Protocol (MPP) has surged in recent months. The number of transactions has crossed 2 million during several periods in late August and early September. Infact, daily agentic-payment volume approached $70,000 at its peak in the period shown on Artemis' three-month dashboard.

The growth comes as AI agents increasingly gain the ability to interact directly with financial infrastructure. They are now executing transactions through APIs, wallets and trading platforms. But where are these agents actually trading?

Coinbase's latest data offers a snapshot of which AI interfaces are actually driving this activity.

Which is the best model for Agentic Trading?

As per the data, Grok generated 25.31% of notional agentic trading volume last week, making it the largest named AI client on the platform. Perplexity followed with 21.61%, while Claude accounted for 8% and ChatGPT for 6.13%.

The ranking has changed significantly in just one week.

Previously, Perplexity accounted for 26.8% of agentic trading volume, followed by Claude at 19.3% and Grok at 15.3%. ChatGPT represented only 1.1%.

Grok therefore gained almost 10 percentage points in a week, while ChatGPT's share increased by more than five times.

Coinbase attributed Grok's jump to its Coinbase for Agents x Grok integration, which pushed the AI model to the top of the latest leaderboard.

The data also shows that agentic trading is not yet dominated by the major consumer AI models.

Custom CLI and other interfaces accounted for 38.47% of notional trading volume last week, making them the largest category. In the previous dataset, the equivalent CLI/Other category represented 36.8%.

This suggests developers are increasingly building customized agentic trading workflows rather than relying solely on off-the-shelf AI assistants.

AI models become trading interfaces

The shift represents a change in how users can interact with crypto markets.

Traditional trading requires a user to open an exchange, analyse an asset, decide on an order and execute the transaction. Agentic systems can potentially compress those steps into an automated workflow in which an AI model interprets a user's instructions and uses connected financial infrastructure to execute the required action.

Coinbase has been building infrastructure specifically for this transition.

Its Coinbase for Agents platform allows AI agents to connect with Coinbase accounts and perform activities including trading and payments through tools designed for agentic applications.