By Krystal Hu
(Reuters) -Harmonic, an artificial intelligence startup co-founded by Robinhood CEO Vlad Tenev, has raised $120 million in new funding, valuing the company at $1.45 billion, as it tackles AI “hallucinations” — or incorrect or nonsensical answers — by improving the ability to reason.
The Series C round for the pre-revenue startup was led by Ribbit Capital, with participation from existing investors Sequoia and Kleiner Perkins. Laurene Powell Jobs’ investment firm Emerson Collective joined as a new backer.
The deal marks the company’s third major fundraising in 14 months, bringing its total capital raised to $295 million, highlighting strong investor interest in startups trying to make AI more accurate and reliable, even before they have commercial products.
Harmonic is developing what it calls “Mathematical Superintelligence” (MSI), a form of AI focused on advanced reasoning that it claims is free of hallucinations and other factual errors that plague many generative AI models.
It says its flagship model, Aristotle, trained on synthetic math proofs — computer-generated examples used to teach problem-solving — achieved a top-level performance at the International Mathematical Olympiad in July alongside Google and OpenAI, a win that CEO Tudor Achim said helped attract investor interest.
Founded in 2023, Harmonic says it can achieve this by using formal reasoning, requiring its AI to output its reasoning as computer code in the Lean4 programming language, which can be checked for correctness. The bulk of the new funding will go toward the immense computing power required for training its models, according to Achim.
By focusing on verifiable, error-free logic, Harmonic says it aims to build trust for AI in safety-critical industries like aerospace and finance, where mistakes can have severe consequences.
“The elimination of hallucinations comes directly from the fact that we require our system to output reasoning as code instead of reasoning as English,” CEO Tudor Achim said in an interview.
Harmonic currently offers its Aristotle model to the public via a free API, a tool that lets developers plug the model into their own software. The company said mathematicians and researchers have been using the tool to check complex proofs and accelerate novel discoveries. Achim said it will explore commercialization in the future.
“I think there are certain areas of software development where safety and reliability are paramount,” he said, adding there is also demand from safety-critical sectors such as automotive and aerospace.
(Reporting by Krystal Hu in San Francisco; Editing by Richard Chang)
Comments