OpenAI model solves open Erdős math problem; Google DeepMind releases Gemini 3.5 Flash.
OpenAI's autonomous resolution of the 1946 planar unit distance problem marks a critical inflection point for AI in formal reasoning and pure mathematics. Moving beyond heuristic pattern matching, this demonstrates a general-purpose model generating novel, verifiable mathematical constructions that outperform human intuition. Meanwhile, Gemini 3.5 Flash's release signals continued rapid iteration in the highly competitive lightweight model tier.
OpenAI has announced a landmark achievement in artificial intelligence and pure mathematics: a general-purpose reasoning model has autonomously solved the planar unit distance problem, a famous open question posed by Paul Erdős in 1946. Concurrently, Google DeepMind announced the release of Gemini 3.5 Flash, continuing the rapid cadence of lightweight, high-efficiency model deployments.
Technical Details The planar unit distance problem asks for the maximum number of unit distances that can exist among a set of n points in a 2D plane. For decades, mathematicians believed the optimal arrangements were rooted in square grid structures. OpenAI's reasoning model overturned this assumption by discovering an entirely new family of geometric constructions that strictly outperform the square grid bounds. Notably, this was achieved by a general-purpose reasoning model rather than a narrow, specialized algorithmic solver, marking the first time such an AI has resolved a prominent, central open problem in mathematics.
In parallel, Google DeepMind's release of Gemini 3.5 Flash introduces the next iteration of their architecture optimized for speed and cost-efficiency. This ensures the frontier lab race remains highly competitive at the low-latency, high-throughput tier.
Why It Matters From an engineering perspective, the OpenAI breakthrough is a massive signal that AI reasoning is crossing the threshold from sophisticated pattern interpolation to true zero-shot extrapolation and novel discovery. Generating a mathematically rigorous proof and a novel geometric construction requires deep, multi-step logical coherence without hallucination—historically a major failure mode for neural networks. This validates the scaling laws of inference-time compute and self-play in formal logic domains.
What to Watch Next Engineers and researchers should monitor for the release of the formal proof and the specific methodology OpenAI used to verify the model's output (e.g., integration with formal proof assistants like Lean). Additionally, watch for API access, context window limits, and pricing details for Gemini 3.5 Flash to evaluate its latency and cost-to-performance ratio against competitors like GPT-4o-mini and Claude 3.5 Haiku.