Performances
RD Popularity Index
RD Growth Index
Comparison
Follow Updates
Follow Lists
People
About Fraction AI
Fraction AI is a decentralized AI model training platform that allows users to create and train AI agents. Its core business enables users to define agent strategies in natural language, compete with other agents in "spaces," where AI judges score their outputs, and winners receive USDC rewards and Fractals points. The platform uses QLoRA technology to fine-tune top-performing agents, forming an automated training loop. Users fully own their agents, control their evolution, and earn from their wins.
The core competitiveness lies in its decentralized training model, which breaks the monopoly of traditional AI training by large corporations and allows individuals to participate and benefit. Additionally, the decision-making and training processes of agents are transparent and auditable, enhancing trust.
Market pain points include: high costs of traditional AI training, centralized data, inability for users to participate or share profits, and lack of transparency and verifiability. Fraction AI addresses these by offering an open platform where users can create agents, participate in competitions, earn rewards, and solve these issues.
In the past six months (February to August 2026), Fraction AI has made significant progress. In February, the platform announced a partnership with Grvt to bring AI agents into on-chain trading infrastructure, launching Signature Agents that allow users to define stablecoin management strategies and deploy them into DeFi protocols. This feature integrates Coinbase's x402 protocol for no-code automated trading. As of August 2026, the platform has over 192,500 users, deployed 152,400 agents, and completed 3.7 million sessions. Additionally, the platform launched the Stable-Up protocol, managing over $100 million in assets with an approximate 7% annual yield. These developments demonstrate the platform's expansion from pure training competitions to real-world financial applications.
Fraction AI, an AI Web3 platform for model training, began with a $6 million pre-seed round in December 2024, led by Spartan and Symbolic, with participation from Polygon and NEAR co-founders. This funding underpinned its R&D and testnet rollout. In early 2025, the testnet v0.1 launched, attracting over 320,000 users who created 1.1 million agents and generated 30 million data sessions, proving the viability of its decentralized reinforcement learning framework. The mainnet went live on Base in May 2025, transitioning from testnet to scalable deployment. This milestone enabled real-world AI agent competitions across domains like copywriting, coding, and finance, with rewards in FRAC tokens. These events mark a clear trajectory: securing capital, validating the protocol, and launching a production network—each step reinforcing its long-term goal of democratizing AI training through permissionless, user-driven feedback loops.
Fraction AI has advanced significantly in H1 2026. The launch of Signature Agents marks a key milestone, enabling users to create and share personalized AI agents for stablecoin management via natural language. This expands the Stable-Up protocol into a user-driven ecosystem. A major partnership with Grvt integrated decentralized AI agents into on-chain trading infrastructure, allowing automated strategies and capital deployment into vaults like GLP. Stable-Up remains the core product, with agents tracking APYs and rebalancing across protocols. Official roadmap focuses on expanding Signature Agents access beyond FOXX NFT holders through the Stable-Up app, enhancing agent interoperability with payment protocols like x402, and growing the ecosystem around transparent, on-chain performance tracking.
Shashank Yadav (Founder): Ex-Goldman Sachs ML Strategist, Microsoft intern, IIT Delhi graduate. Led Fraction AI's product and strategy.
Rohan Tomar (Co-Founder): IIT Delhi graduate, previously at Goldman Sachs and Microsoft. Focused on technology and operations.
Charles Elkan (Core Contributor): Renowned ML researcher (UCSD professor), advising Fraction AI's data labeling methodology.