Wallet V, a self-custody Web3 wallet incubated by Virgo Group, has launched a public performance benchmark for AI trading agents configured by its users on the decentralized derivatives platforms Hyperliquid and Aster. The benchmark, hosted on the Wallet V website, aggregates the performance of 688 agents created over the prior two months, offering transparency into how different large language models (LLMs) perform as autonomous trading engines.
Each agent was set up by a user who selected an LLM to generate trading decisions, and the agent then executed perpetual futures strategies on either Hyperliquid or Aster. The aggregated data, refreshed as new agents are deployed, spans seven LLM families. 42 percent of the agents recorded a break‑even or positive profit‑and‑loss balance over the period. The best‑performing model delivered a peak return on investment of 307 percent, while the weakest model posted a 30 percent loss. For model families represented by fewer than 10 agents, the results are reported as directional only.
The agents traded across four asset classes available on the two platforms: major digital assets such as BTC, ETH, and SOL; pre‑IPO equity exposure; commodity benchmarks including gold, silver, and oil; and major foreign exchange pairs. All instruments are accessed through third‑party venues, with Wallet V acting solely as the connecting software.
“At Wallet V, the focus has been on building infrastructure for the next phase of crypto. This benchmark is what that next phase looks like up close. Users now decide which AI model to configure their agent in the same way institutions evaluate managers, by reviewing observable performance over time,” said Adam Cai, Founder & CEO of Virgo Group.
Wallet V plans to extend the benchmark with newer model families, support for prediction markets, advanced copilot trading analytics, and personalized AI prompt generation tailored to individual trading styles. The wallet application is available for iOS and Android at dl.walletv.io.
