OpenAI has introduced EVMbench. This is a new tool that allows you to measure the performance of artificial intelligence agents in a series of tests (benchmarks) by discovering, modifying, and leveraging smart contracts on Ethereum. The company developed EVMbench in collaboration with Paradigm, a company that funds cryptocurrency network projects.
OpenAI AI bots on Ethereum can evaluate three important aspects of the protocol.
beginning, Actual vulnerability detected in Ethereum open source Uses data from public audits. We then assess the risk of vulnerabilities and audit rewards for the rest of the developers in our ecosystem.
Number 2, Suggest patches to fix these vulnerabilities without compromising the operation of the protocol. “Agents must modify vulnerable contracts to eliminate exploitability while preserving intended functionality. This is verified through automated testing and exploit checks,” OpenAI said.
Third, Simulate attacks that exploit these flaws to exfiltrate funds Controlled simulation environment (sandbox) Safe. However, OpenAI clarifies that EVMbench “does not fully represent the complexity of real-world smart contract security.”
EVMbench uses 120 real-world vulnerabilities extracted from 40 public competency audits, including Code4rena. Performance tests conducted so far show that the GPT-5.3-Codex agent model achieved a 72.2% success rate in exploiting the flaw. The previous model’s GPT-5 success rate was only 31.9%.
However, vulnerability detection and patching The results of the model are not very encouraging.
Performance is degraded for discovery and patching tasks. During the discovery phase, the agent may stop after identifying a single issue rather than thoroughly auditing the codebase. During the patching phase, maintaining full functionality while eliminating subtle vulnerabilities remains a challenge.
OpenAI, an artificial intelligence company.
Why is EVMbench important?
According to the company, the importance of this agent audit tool lies in the fact that smart contracts routinely guarantee $100 billion in crypto assets within open source protocols.
“As AI agents improve their skills in reading, writing, and executing code, it becomes increasingly important to measure their capabilities in economically appropriate environments and encourage the use of AI systems defensively to audit and enforce deployed contracts,” the company said in a statement at the product presentation.
OpenAI agents on Ethereum come at a time when autonomous agents are rapidly advancing within the crypto asset ecosystem. As reported by CriptoNoticias, these are already able to interact with complex environments such as the Lightning Network. In this second layeragents can manage liquidity channels and economic interactions with other AIs.
Coinbase has launched Agentic Wallet, a wallet that allows AI agents to operate on the Base network without paying fees. And Phantom, Solana’s most popular wallet, activated its MCP server to allow AI agents to manage balances and operate autonomously.
According to data from Token Terminal, the number of weekly transactions on the Ethereum network reached 17.3 million. Explosion of transactions on the network Occurs after the launch of ERC-8004 in August 2025. It is a standard that enables “discovering, selecting, and interacting with agents across organizational boundaries without the need for pre-existing trust” in the open agent economy.
(Tag translation) Smart contract
