Cyber, EigenLayer, Sentient, RootData and others jointly launched Crypto AI Benchmark Alliance to set a new benchmark for Crypto AI

14 blockchain and artificial intelligence projects including Cyber, EigenLayer, Sentient jointly announced the establishment of Crypto AI Benchmark Alliance (CAIBA) . This open source, community-driven alliance will be committed to establishing transparent and trusted evaluation standards for AI models and agents in the crypto industry.

The first initiators - Alchemy, Cyber, Dune, EigenLayer, Goldsky, IOSG, LazAI, Magic Newton, Metis, MyShell, OpenGradient, RootData, Sentient and Thirdweb -will work together to contribute data sets, tools and expertise to jointly build an evaluation framework. Each set of benchmarks will include tasks, reference answers and scoring scripts, and will be published on platforms such as GitHub and Hugging Face under an open license (when the scope of the license allows).

As AI applications in the crypto field continue to expand, from trading strategies to research assistants to all-inclusiveness, traditional AI benchmarks have no longer reflected the industry's unique needs. CAIBA aims to fill this gap and launches professional reviews for encryption scenarios.

“Transparent and rigorous testing is crucial,” said Ryan Li, co-founder of Cyber. “Models must not only answer the questions correctly, but they must also be executed reliably, giving users more confidence in their decisions.”

The first achievement of the alliance a Benchmark for Crypto AI Agents (CAIA) is now online, which measures AI capabilities from three dimensions:

• Knowledge : Accurately answer questions such as protocols and tokens
• Planning : Develop a multi-step task plan
• Action : Complete operations with the help of a block browser and API

CAIA covers scenarios such as token economics, on-chain analysis, project research and trading processes. The evaluation objects include general big models such as GPT-4o, Claude 4, Gemini 2.5, and DeepSeek-R1, as well as a variety of cryptographic native models.

By testing models in real tasks, CAIBA has established a unified and reproducible measurement standard for encrypted AI, helping the industry build more trustworthy intelligent applications. The Alliance is already developing more benchmarks and welcomes new members to join. Developers, researchers and protocol teams can submit models to participate in evaluation or propose brand new tasks.

---

About Crypto AI Benchmark Alliance (CAIBA)

Crypto AI Benchmark Alliance is an open alliance for community governance that focuses on formulating AI evaluation standards for crypto scenarios. Through open datasets, reproducible tasks and public rankings, CAIBA provides developers, researchers and protocols with tools to measure and improve AI systems in blockchain applications. For more information, visit caiba.ai .