What is GPT-4?

Giving away GPT-4 for free was both revenue positive and retention positive for OpenAI, contradicting the conventional wisdom to keep premium model tiers behind a paywall.

“We just gave it away for free and that ended up being totally revenue positive and retention positive.”

GPT-5 launched at a lower price point than GPT-4, a result that surprised beta testers and runs counter to assumptions that frontier model releases command premium pricing.

“For five to be cheaper than four, I don't think anyone expected.”

Reiner Pope on why context lengths have stalled at 100-200K tokens.

“If you look at the history of context lengths of models, from earlier models like GPT-3, maybe to GPT-4, I don't remember when the transition happened exactly, they shot up from about 8K to 100-200K. And then for the last year or two, they've all been hovering around there. I think that indicates that this is the reasonably balanced cost point, and going massively beyond that would be cost-prohibitive. Not because of the compute cost, because of the memory bandwidth... Because of memory bandwidth cost, yeah.”

DeepSeek delivered GPT-4-level performance at 1/600th the inference cost of GPT-4.

“Deep Seek, for example, on GPT-4 was 1/600 the cost.”

Joseph Nelson on the state of computer vision relative to language AI maturity.

“Computer vision today is roughly where language capabilities were 3 years ago with the introduction of ChatGPT and GPT-4.”

GPT-4 passing false belief tests does not predict generalization to real-world tasks requiring genuine inference of human intent.

“We should be concerned about how well this will generalize to real tasks where we might care about this much more deeply.”