What is IMO?

GPT (version 5.6 as of 2026) autonomously produced a counterexample to the Erdos unit-distance conjecture and now achieves perfect IMO scores on demand, marking it as a frontier tool for serious mathematical research.

“The fact that I can push a button on GPT 5.6 and get perfect scores on the IMO.”

GPT (OpenAI frontier models) autonomously disproved the Erdos unit distance conjecture and achieved perfect IMO scores, signaling a shift from benchmark-chasing to open-ended mathematical discovery.

“It turns out that the conjectured value is just wrong. And, autonomously, GPT found a counterexample, you know, constructed a solution where this exponent is not satisfied.”

Alex Kontorovich on AI math benchmarks becoming obsolete via GPT 5.6 and IMO perfect scores.

“The fact that I can push a button on GPT 5.6 and get perfect scores on the IMO.”

Kontorovich explains why AI companies will stop entering math olympiad competitions: once frontier models can already score perfectly with a button press, competitive entry holds no interest or marketing value for them.

“My prediction was that this summer, no AI models would get a gold medal 'cause none of them would bother competing anymore. So it's not interesting for them. The fact that I can push a button on GPT 5.6 and get perfect scores on the IMO.”