Interconnects: Open and closed models are on different exponentials
Source
Nathan Lambert
Author
Nathan Lambert
Date
Key takeaways · AI-distilled
The whole open-versus-closed question reduces to one economic bet: will users keep paying a large premium for the best model? Coding agents are the first market to answer yes — the author says he would pay $2,000/month, because net output visibly rises.
Closed labs win on returns to integration: weights, agent harnessThe scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.Full definition →, tools, and serving infrastructure are co-designed. Open models must run across many diverse serving setups, so they cannot capture those gains, and every layer having alternatives drives prices to commodity.
Expect frontier labs' API businesses to decay on purpose. Labs will hold their best models back from APIs to protect scarce tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → supply, prevent distillationTraining a small, cheap model to imitate a big one's outputs, keeping much of the capability at a fraction of the cost.Full definition → by competitors, and keep capacity aimed at higher-margin uses.
benchmarkA standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.Full definition → saturation is not the ceiling people think it is. If absolute intelligence tops out, labs shift to optimizing utility per second or per watt — improvement continues, it just stops showing up in the scores everyone watches.
Enterprises buy open models differently: they pick a model that clears a threshold on one task and then never swap it, because setup cost dominates. Value accrues across many vendors rather than one, which is why the open market can be larger while margins are thinner.
Terms in this piece · Glossary
agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
distillation — Training a small, cheap model to imitate a big one's outputs, keeping much of the capability at a fraction of the cost.
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
Why it matters
The case that open and closed models sit on different capability exponentials — and what that means for betting on either.
Key quotes
“For people who rely on coding agents to work, they will always pay more for the best rather than settle for good enough.”
“In 5-10 years I expect both OpenAI and Anthropic to be valued in the $2-10T range.”
“The true frontier labs will be an oligopoly that looks like the cloud market today.”
“The total market value will dramatically exceed the cumulative value of OpenAI and Anthropic.”
“The key is that the open and closed model economies are operating on different exponentials.”
