Wallet V launched a public performance benchmark tracking 688 AI trading agents across Hyperliquid and Aster platforms over two months. Forty-two percent of agents achieved break-even or profitable results, with returns ranging from a 30% loss to a 307% gain. The report covers strategies executed on major digital assets, pre-IPO equities, commodities, and forex pairs.
688 Agents Across Seven Model Families
The benchmark aggregates data from user-created AI agents built on seven large language model families. Most agents traded perpetual futures contracts, with performance measured across BTC, ETH, and SOL digital assets alongside gold, silver, oil commodities, forex pairs, and pre-IPO equities. Models represented by fewer than ten agents received directional performance indicators rather than precise metrics.
Extreme Performance Range Observed
While 42% of agents ended with non-negative profit and loss, individual returns showed massive volatility. The highest recorded gain hit 307%, but losses reached 30% at the lower end. This wide spectrum highlights how small changes in model parameters or market timing dramatically impacted outcomes for identical asset classes.
Wallet V's Third-Party Connection Model
Wallet V functions as a Web3 self-custody wallet that links users to external trading platforms but provides no trading services or AI tools directly. The company serves as an access layer while third parties handle execution. Virgo Group, led by CEO Adam Cai, backs Wallet V with investment from Draper Dragon, OKX Ventures, and Waterdrip Capital.
Planned Benchmark Expansions
Wallet V confirmed upcoming additions including support for newer large language model families, prediction markets integration, and personalized AI prompt generation features. The next benchmark iteration will incorporate these enhancements while maintaining the two-month performance tracking structure. The company hasn't set a release date for the expanded framework.
