“if they put them all in one cluster, they could have a bigger model than any of the labs next year. Right?”
Patel reports AI models improved from below 4% to 70% accuracy on Frontier Math in six months across multiple companies.
“frontier math. And the models went from sub 4% to 70% in six months. Right? And this is from not just OpenAI, but also Anthropic and Google and DeepSeek.”
Patel reports AI models improved from under 4% to 70% on Frontier Math in six months.
“Like, can't do all the math, all the problems in frontier math. And the models went from sub 4% to 70% in six months.”
Patel explains OpenAI hides reasoning chains while Anthropic shows them, creating distillation advantage for closed models.
“You go look at like codex or clawed code, you give it a question, it thinks for a while, and then it gives the answer.”
Forecasting the future gap between open and closed AI models requires measuring historical performance differences.
“To project how the open vs closed capability gap will progress in the future, we first need to measure the past.”
An unnamed company generates more revenue from AI model usage than Anthropic does.
“They’re making way more money off of using these models than Anthropic.”