Fraction AI, a crypto-AI startup focused on decentralizing data labeling, has raised $6 million in a pre-seed funding round co-led by Spartan Group and Symbolic Capital.

Other investors in the round include Borderless Capital, Anagram, Foresight Ventures and Karatage, Fraction AI said Wednesday. Angel investors, including Sandeep Nailwal of Polygon and Illia Polosukhin of NEAR Protocol, also joined the round and are "close advisors" to the project, Fraction AI founder and CEO Shashank Yadav told The Block.

Fraction AI began raising funds for the pre-seed round in April of this year and closed it in September, Yadav said. The round was structured as a simple agreement for future equity (SAFE) with token warrants, Yadav said, declining to comment on the post-round valuation.

Founded in February, Fraction AI is a crypto-AI startup aiming to decentralize data labeling. Data labeling involves tagging raw data —such as images, text or audio — with meaningful labels to train AI models to recognize patterns and make accurate predictions.

"Among AI's three core elements — data, compute and models — data remains the most obscure and tightly controlled," Yadav said. "We set out to change that, leveling the playing field and empowering anyone to train high-quality AI models."

Fraction AI follows a hybrid approach to data labeling, combining human insights with AI agents. The platform will have three main participants: stakers, builders and judges.

Stakers will earn rewards by staking ether (ETH) or liquid staking tokens (LSTs) like Lido staked ether (stETH). Their earnings will come from entry fees paid by builders, with 5% of each entry fee going directly to stakers, Yadav said.

Builders will create agents by providing human insights or detailed instructions in text form with no coding required. They will fund their agents with ETH or LSTs, enabling them to compete in challenges. Builders will pay a small entry fee to participate and generate the best possible data. The top three agents in each competition out of five will be rewarded from the entry fee pool, with their rewards multiplied by a performance-based multiplier, determined by scores from specialized large language models (LLMs).

"The additional multiplier-based reward comes from the stakers' pool, ensuring poor performers fund stakers while exponentially boosting rewards for top-performing agents," Yadav said.

Judges, which are specialized LLMs, will evaluate agent outputs against predefined criteria. To participate, judges will need to stake Fraction AI’s native FRAC tokens.

Fraction AI is primarily built on Ethereum and is currently live on a closed testnet with over 60,000 users, Yadav said. The public testnet is expected to launch next month, with the mainnet slated for release by the end of the first quarter or early second quarter of 2025.

The FRAC token will also be launched closer to the mainnet, Yadav said. The token's use case will be to secure a network of judges who will evaluate agent outputs through staking and slashing mechanics, ensuring quality and fair evaluation, he added.

While Fraction AI is currently primarily built on Ethereum, it plans to launch on NEAR as well as multiple Ethereum Layer 2 networks, Yadav said.

The project, based in San Francisco, currently has eight employees. Yadav plans to keep the team lean for the near future.

Crypto-AI startups continue to attract VC funding, as The Block reported recently. Skepticism persists, however, with investors warning that many such startups could fail as they chase trends without clear value propositions.

The Funding newsletter:Stay updated on the latest crypto funding news and trends with my bimonthly newsletter, The Funding. It's free. Sign uphere!

Fraction AI Raises $6 Million in Pre-Seed Funding to Decentralize Data Labeling