LLM comparison: the four models we test, in facts and builds

Most LLM comparison pages line up context windows, prices and benchmark scores. Those numbers are here too, but they are the footnote, not the verdict. The reviews on this shelf judge each model by what it built from the same briefs as the others, one attempt, recorded with the same script.

Prices were checked on October 21, 2026, on Anthropic's model overview and OpenAI's GPT-6.1 Sol announcement. All four models list a context window of around a million tokens, so context size does not separate them on briefs this size.

What does separate them is visible on every model page: how long each one works, how much it writes, and what the result looks like when nobody tidies it up. The reviews below add release context, the vendor's own claims, and an honest read of where those claims hold and where the builds disagree with them.