In a boost for decentralized AI projects, BitAgent (Bittensor subnet 20) has grabbed the number one spot on a key industry ranking for how well small AI systems can handle real-world tasks.
The Berkeley Function Calling Leaderboard (BFCL), run by researchers at UC Berkeley, tests AI models on their ability to accurately “call functions,” which basically means giving clear instructions to software tools to get jobs done, like automating routine work. Think of it as measuring how reliably an AI can act as a helpful assistant without messing up.
BitAgent SN20, developed through a community-driven system on Bittensor where participants (called “miners”) compete to improve models, topped the chart for AI systems with under 8 billion parameters—a measure of model size that keeps them efficient and affordable to run. It scored 46.6% overall accuracy, edging out rivals from big tech firms like Salesforce’s xLAM model at 46.56% and others trailing behind.
This win shows how decentralized networks, which reward innovation through crypto incentives, can challenge centralized giants like Microsoft and Anthropic in building practical AI.
For everyday users, this matters because BitAgent powers tools for IT companies, such as MSPTech.ai, to automate services like fixing tech issues or managing networks. Industry experts predict that AI like this could cut operational costs by up to 30% in the coming years by handling more tasks on its own.
As Bittensor’s ecosystem grows, successes like this could speed up the shift toward smarter, more automated business operations.
Check the announcement tweet:
Be the first to comment