Wallet V Publishes Public Benchmark for AI Trading Agents on Hyperliquid & Aster
Wallet V, a self-custody Web3 wallet, has launched a public performance benchmark for AI trading agents configured by its users on the decentralized derivatives platforms Hyperliquid and Aster. The benchmark is hosted on Wallet V’s website and publishes aggregated cohort results.
Wallet V analyzed 688 AI trading agents created over the prior two months. Each agent was user-configured, used a user-selected large language model to generate trading decisions, and executed trades on Hyperliquid or Aster. Wallet V aggregates performance by underlying model family, and refreshes results as new agents are deployed.
Key stats from the cohort: 42% of agents achieved a profit/loss balance at or above zero. Peak agent-level ROI ranged from -30% (lowest-performing model) to +307% (highest-performing model). Model families with fewer than 10 agents are treated as directional rather than statistically conclusive.
Across the cohort, agents traded perpetual futures across four asset-class buckets: major crypto assets (BTC, ETH, SOL), equity exposure (including pre-IPO exposure), commodities benchmarks (gold, silver, oil), and major FX pairs. Instruments are accessed via third-party venues.
Adam Cai (Founder & CEO of Virgo Group) said the benchmark gives users an institution-like way to evaluate models based on observable performance. Wallet V plans further releases, including additional model families, prediction market support, advanced analytics for copilot trading, and personalized AI prompt generation.
For traders, the release adds a new public dataset to compare AI trading agents and model families, but performance dispersion (from -30% to +307% ROI) suggests results may be highly regime-dependent.
Neutral
The launch is primarily a data/benchmarking upgrade rather than a direct liquidity or token-issuance event. By publishing aggregated results across 688 AI trading agents and showing wide dispersion in ROI (from -30% to +307%), Wallet V may improve trader confidence in how to select models, but it also highlights that AI trading performance can vary sharply by model family and market regime.
In the short term, traders may lean toward monitoring the benchmark when evaluating AI-driven strategies on Hyperliquid/Aster, potentially increasing attention and usage of specific model families. In the long term, if the dataset proves to be predictive and the refresh cadence remains consistent, it could gradually legitimize AI-agent evaluation in onchain derivatives.
Because the news does not indicate new token incentives, protocol changes, or guaranteed performance, the net effect on broader market stability is likely limited—hence a neutral stance rather than clearly bullish or bearish.