OpenAI's "28 Days of Codex" campaign has fueled widespread speculation that the event will culminate in a new model launch. Rather than the flagship reasoning upgrades or cheaper lightweight options developers have been requesting, rumors increasingly point toward a model dubbed "GPT-6.1 Terra" - a prospect that has generated more unease than excitement in developer communities.
How OpenAI's Model Tiers Are Structured
OpenAI's lineup is generally understood to split across three tiers. Lightweight models, referred to in developer discussions as the "Luna" class, prioritize speed and low cost for high-volume tasks like background API calls and simple conversational agents. At the opposite end, top-tier "Sol" and "Astra" class models deliver the deep reasoning, multi-step logic, and code comprehension needed for complex engineering work. In between sits "Terra," a hybrid architecture designed to balance speed and reasoning power.
In practice, developers say that balance rarely materializes. Mid-tier models in this category tend to be too slow and expensive to compete with budget options for high-concurrency workloads, while lacking the consistency and depth required to match flagship models on demanding reasoning or multi-turn coding tasks.
Why Developers Wanted Something Different
The Codex campaign was framed around improving programming assistance and developer tooling, and community expectations coalesced around two priorities. The first is further refinement of flagship reasoning models - improvements or cost reductions that make long-context refactoring and zero-shot coding more dependable. The second is efficiency gains for budget-tier models, including expanded context windows that let smaller teams and independent developers scale without facing steep API costs.
An incremental update to a mid-tier model does not clearly address either priority, which is why speculation about "GPT-6.1 Terra" has been met with skepticism rather than enthusiasm among developers tracking the countdown.
What Remains Uncertain
It is important to note that these details are based on community speculation and forum discussion rather than an official OpenAI announcement. The naming conventions "Luna," "Terra," "Sol," and "Astra" reflect how developers have informally categorized OpenAI's offerings, not necessarily the company's own branding. Whether OpenAI ultimately prioritizes flagship reasoning improvements, budget-tier efficiency, or a different release entirely will only become clear once the 28-day campaign concludes and an official announcement is made.