GPT-5’s model lineup, posted by GitHub before OpenAI was ready — confirmed

GitHub briefly announced part of OpenAI’s GPT-5 lineup before OpenAI’s own presentation on 7 August 2025. The Verge reported a quickly removed GitHub blog post describing several versions of the model and promoting improvements in reasoning, code quality and agent-like work. The disclosure came from a partner’s publishing system, not an anonymous claim about a secret benchmark.

That made the names and intended availability more substantial than the usual launch speculation. It did not independently validate GitHub’s performance language. A prematurely published product announcement is still marketing material, and stronger coding or reasoning claims require testing rather than a recognizable company logo.

The Verge recorded how the post had been noticed and removed. Other launch-day coverage followed when OpenAI made GPT-5 public. TechCrunch and Engadget described the actual rollout, the different offerings and the effort to simplify ChatGPT’s model selection. Those reports provide the later check on the existence of the lineup exposed by the partner post.

The confirmed badge is scoped to the premature announcement and the subsequent launch. It is not a rating of the product or an endorsement of predictions that GPT-5 would transform particular industries. The difference matters because release-day excitement can turn an accurate product leak into an excuse to repeat much broader, untested promises.

The practical lesson is familiar to anyone who coordinates a large launch: publication across partners creates more places for a secret to escape. This archive links contemporary reporting and launch coverage rather than an archived copy of the removed blog. We did not verify a suitable X post showing the early GitHub disclosure, so there is no fabricated embed or engagement count. Later launch posts were available, but substituting one would misrepresent where this particular story broke.