Skip to main content

Logical Intelligence

Category: Foundation Models / LLMs

A frontier AI research lab developing non-autoregressive, energy-based reasoning models (EBMs) designed to achieve mathematical certainty and logical consistency beyond the capabilities of traditional LLMs. Logical Intelligence was founded in 2025. The company is led by Eve Bodnia. Based in San Francisco, USA. Team size: 10-50. Total funding raised: Undisclosed. Latest round: Series A. Key investors include ["Undisclosed (High-profile)"].

Founded
2025
Headquarters
San Francisco, USA
Team size
10-50
Total funding
Undisclosed

Value proposition

Eliminates the 'probabilistic guessing' and hallucinations inherent in autoregressive token prediction by using energy-based constraints to ensure verifiable reasoning and formal correctness.

Products and solutions

["Kona 1.0 (Energy-based reasoning engine)","Aleph (Internal mathematical discovery and benchmarking tool)","Token-free General Reasoning Framework","Formal Verification Suite for Code and Mathematics"]

Unique value

The company is one of the few frontier labs successfully pivoting away from the industry-standard Transformer/Autoregressive architecture toward Energy-Based Models (EBMs), led by a 'dream team' of Turing and Fields Medal winners.

Target customer

Enterprise developers of critical systems, formal verification engineers, academic research institutions, and industries requiring high-stakes mathematical accuracy (e.g., aerospace, cryptography, and quantitative finance).

Industries served

["Artificial Intelligence","Software Engineering (Formal Methods)","Mathematics & Scientific Research","Quantum Information Science","Cybersecurity"]

Technology advantage

Proprietary 'Kona' architecture allows for reasoning by enforcing mathematical constraints rather than predicting the next token; demonstrated superior performance on the PutnamBench (76% accuracy), significantly outperforming standard LLMs in formal logic.

How they differentiate

Utilizes non-autoregressive Energy-Based Models (EBMs) to enforce mathematical constraints and logical certainty, eliminating the 'probabilistic guessing' and hallucinations common in traditional Transformer-based LLMs.

Main competitors

["OpenAI","Google DeepMind","Harmonic"]

Key partnerships

["Harvard Center of Mathematical Sciences and Applications (CMSA)","Meta AI (via Yann LeCun's advisory role)","UC Santa Barbara (Quantum Information research ties)"]

Notable customers

["Harvard Center of Mathematical Sciences and Applications (CMSA)","Formal Verification Engineers"]

Major milestones

["Appointed Yann LeCun as Founding Chair of Technical Research Board in Jan 2026","Recruited Fields Medalist Michael Freedman as Chief of Mathematics","Developed Kona 1.0, the first energy-based reasoning AI engine","Achieved 76% on PutnamBench with the Aleph internal tool"]

Growth metrics

Achieved 76% accuracy on the PutnamBench benchmark; recruited a 'dream team' including a Turing Award winner and a Fields Medalist.

Market positioning

Frontier AI research lab focused on formal verification and high-stakes reasoning for critical systems.

Geographic focus

North America (San Francisco), Global Research

Patents and IP

No registered patents disclosed; currently operating on proprietary research and trade secrets related to non-autoregressive architectures.

About Eve Bodnia

Eve Bodnia is a mathematician and physicist with a PhD in Quantum Information and Algebraic Topology from UC Santa Barbara. She studied under 2025 Nobel laureate Michel Devoret. Bodnia has authored over 20 academic papers on dark matter, quantum mechanics, and particle physics. Prior to founding Logical Intelligence, she conducted research at the intersection of geometry and machine learning and hosted high-level academic workshops at Harvard's Center of Mathematical Sciences and Applications (CMSA).

Official website: