Artificial intelligence startup Fetch.ai has launched a new native Web3 applicable large language model (LLM) designed to support agentic AI.
The release of ASI-1 Mini on Tuesday integrates it into Web3 ecosystems, enabling secure and autonomous AI interactions. This launch lays the foundation for broader innovation in the AI sector, paving the way for the imminent debut of the Cortex suite, which will further enhance large language models and generalised intelligence.
“This launch marks the beginning of ASI-1 Mini’s rollout and a new era of community-owned AI,” said Humayun Sheikh, CEO of Fetch.ai and Chairman of the Artificial Superintelligence Alliance.
“We’ll soon introduce advanced agentic tool integration, multi-modal capabilities, and deeper Web3 synergy to enhance ASI-1 Mini’s automation capabilities while keeping AI’s value creation in the hands of its contributors.”
Fetch.ai aims to democratise foundational AI models, enabling the Web3 community to use, train and own proprietary LLMs like ASI-1 Mini.
By decentralising AI, Fetch.ai creates opportunities for individuals to directly benefit from the economic growth of cutting-edge models, which could reach multi-billion-dollar valuations.
Through its platform, Fetch.ai allows users to invest in curated AI model collections, contribute to their development, and share in the generated revenues. For the first time, decentralisation is driving AI model ownership, ensuring a more equitable distribution of financial benefits.
Decentralization in AI enhances transparency, security, and accessibility by distributing control across a broader network rather than central entities. It empowers individuals to own, train, and monetize AI models. Subsequently, this fosters innovation while reducing risks of bias and censorship. This approach ensures AI’s benefits are more equitably shared across society.
Read more: D-Wave Quantum sells world’s largest quantum computer to German research firm
Read more: Argentine President Javier Milei in hot water over prospective crypto rug pull
Flexibility allows for a balance depth and precision
ASI-1 Mini enhances decision-making adaptability with four dynamic reasoning modes: Multi-Step, Complete, Optimised, and Short Reasoning.
This flexibility enables it to balance depth and precision based on the specific task, whether solving complex, multi-layered problems or providing concise, actionable insights. Its Mixture of Models (MoM) and Mixture of Agents (MoA) frameworks further strengthen its versatility, ensuring efficient performance across a range of applications.
By dynamically selecting the most relevant AI models from a suite of specialised options, the system optimises for specific tasks or datasets, maximising efficiency and scalability. This targeted approach is particularly valuable for multi-modal AI and federated learning, allowing seamless adaptation to diverse computational challenges.
Independent agents with unique knowledge and reasoning capabilities collaborate to solve complex tasks, while the system’s coordination mechanism efficiently distributes workloads. This approach enables decentralised AI models to thrive in dynamic, multi-agent environments.
The architecture operates across three interacting layers. At its core, ASI-1 Mini functions as the intelligence and orchestration hub. The specialisation layer, known as the MoM Marketplace, houses diverse expert models accessible through the ASI platform.
In the action layer, called AgentVerse, agents manage live databases, integrate APIs, and facilitate decentralised workflows. By selectively activating only the necessary models and agents, the system maintains optimal performance, precision, and scalability for real-time tasks.
ASI-1 Mini delivers enterprise-grade performance on just two GPUs, cutting hardware costs by eightfold while maintaining scalability. It matches or surpasses top LLMs in specialised fields like medicine, history, and business. Furthermore, the upcoming context expansions to 10 million tokens will help handle complex documents. A few examples of which are legal reviews, and large-scale financial analysis.
Read more: Hard times for Bitmain as tariffs and China’s displease with blockchain stall forward momentum
Read more: Sol Strategies chosen as sole staking provider for Solana ETF
ASI-1 offers potential solution for “black box” problem
The AI industry has long struggled with the black-box problem, where deep learning models generate conclusions without clear explanations.
ASI-1 Mini addresses this challenge by using continuous multi-step reasoning to facilitate real-time corrections and optimise decision-making. It does not entirely eliminate opacity, but it does significantly improve explainability. Further, this an essential feature for industries like healthcare and finance.
By leveraging a multi-expert model architecture, ASI-1 Mini enhances transparency while streamlining complex workflows across various sectors. It efficiently manages databases, executes real-time business logic, and outperforms traditional models in both speed and reliability. This design ensures AI-driven processes remain adaptable, precise, and scalable.
ASI-1 Mini will integrate with AgentVerse, Fetch.ai’s agent marketplace. This will equip users with tools to build and deploy autonomous agents through simple language commands.
These “micro-agents” can automate tasks such as trip planning, restaurant reservations, and financial transactions, creating a seamless AI-driven user experience. The platform fosters open-source AI customisation and monetisation, establishing an “agentic economy” where developers and businesses thrive symbiotically.
Developers can generate revenue from their micro-agents, while users gain access to tailored AI solutions designed for real-world applications.
As its ecosystem matures, ASI-1 Mini will be capable of processing structured text, images, and complex datasets with context-aware decision-making.
By advancing decentralised AI and promoting transparency, ASI-1 Mini introduces the prospects for scalable and efficient AI-driven automation across industries.
