Fetch.ai’s ASI-1 Mini Targets Rising Demand for AI Agents in Web3

Key Takeaways

• Web3 agents powered by large language models (LLMs) can autonomously operate across crypto networks.
• Typical features include wallet integrations and specialized knowledge.
• Fetch.ai has billed its new AI model as a “Web3-native LLM” for building agents.

The idea of crypto-native AI, or the agentic web, has been building momentum since 2024, when decentralized AI gained significant popularity.

In the next stage of that journey, fetch.AI on Feb. 25 launched ASI-1 Mini, a “Web3-native large language model (LLM)” built to support “agentic workflows.”

AI Agents in Web3

Among AI systems, “agents” are typically contrasted with more rudimentary  “bots.”

Whereas bots are confined to a limited field or task and typically require human oversight, agents are more capable and autonomous.

Chatbots are being superseded by agents that can use computers and make decisions independently.

Meanwhile, increasingly sophisticated trading agents use similar browser integrations to execute complex strategies across multiple platforms.

In 2024, AI developer Andy Ayrey created a breakout Web3 agent, Truth Terminal, that brought significant attention to the concept.

Truth Terminal (ToT) has its own X account and a crypto wallet. After inspiring the GOAT meme coin and promoting it on X, The AI bot became the first AI millionaire, and the value of its assets surpassed $18 million.

Fetch.ai’s Agent Marketplace

Described as an “open marketplace for AI agents,” fetch.ai is an open-source platform for creating and sharing agents powered by the FET token.

Because it lets users combine different AI models any way they want, there are potentially limitless possibilities for new agent arrangements. In the Web3 arena, popular use cases include compiling on-chain data from different sources, scraping websites and summarizing information.

However, until now, the underlying models that power the agents came from external sources, including open-source models and proprietary ones accessed via API integrations.

ASI-1 Mini

While fetch.ai supports multiple open-source AI models, ASI-1 Mini is the first LLM built and optimized for the platform.

This approach is designed to “democratize access to foundational AI models,” allowing the Web3 community to own them, fetch.ai said in a press release.

Building on the multi-model frameworks developed by fetch.ai users, ASI-1 Mini isn’t really a single model at all.

Rather, it “dynamically selects from multiple specialized models” to deliver the best output for any given task, the press release stated.

What this means is that ASI-1 Mini isn’t just a single AI model but a system that picks the best tool for the job. Instead of relying on one-size-fits-all AI, it uses different specialized models depending on the task.

Other blockchain projects are exploring AI-powered agents, with companies like Alethea AI and SingularityNET also working on decentralized AI solutions.

Meanwhile, OpenAI and Anthropic have hinted at future models with greater autonomy, sparking broader interest in AI agents that can operate independently across different ecosystems.