As of September 22, 2026, the confusion comes from OpenAI mixing three models with different reasoning-effort settings. Those five or six choices are not really five or six independent intelligences.
My recommendation for you
Use Astra Medium as your normal setting.
Move to Astra Extra High when the task is especially consequential or genuinely knotty—for example:
dissecting a MolDx noncoverage rationale;
developing Roche’s TROP2 coding strategy;
reconciling six stakeholder letters;
finding weaknesses in an argument before it goes to a client;
synthesizing conflicting evidence into a defensible policy conclusion.
Use Sol Medium for serious but more bounded work: summarizing documents, researching CPT history, revising a draft, or producing a polished client memo.
You have almost no reason to select Luna manually.
What the names actually mean
| Model | What it is for | Your likely use |
|---|---|---|
| GPT-6 Astra | OpenAI’s most capable model, emphasizing sustained reasoning, research, judgment, and multi-step work | Your difficult analytical and strategic work |
| GPT-6 Sol | Strong general professional model; faster and less resource-intensive than Astra | Routine research, summaries, rewriting, coding/payment lookups |
| GPT-6 Luna | Efficient model for narrow, repetitive, high-volume tasks | Extracting fields, formatting lists, classifying rows |
OpenAI describes Astra as its strongest option for complete workflows requiring sustained reasoning and judgment; Sol for difficult everyday professional work; and Luna for clearly specified, repeatable tasks. (ChatGPT Learn)
Then each model can be given different amounts of thinking:
Light: quick, well-defined work.
Medium: meaningful planning and analysis.
High/Extra High: difficult work involving several steps, sources, or tradeoffs.
Max: the model gets still more time for one extremely hard problem.
Ultra: the task can be divided among multiple agents working on different components.
Thus Astra Medium means “the best model, thinking fairly hard.”
Astra Extra High means “the best model, take your time and really work through this.”
Why you may see exactly six choices
The standard “Power” slider currently contains these six presets:
Luna High
Sol Light
Sol Medium
Astra Light
Astra Medium
Astra Extra High
OpenAI says some plans may omit Astra Extra High. More granular choices—including Max and Ultra—may appear under Advanced or require enabling them in settings. (ChatGPT Learn)
The simple Bruce Quinn rule
Astra Medium: Leave it here by default.
Astra Extra High: “This problem is worth several thousand dollars if handled correctly.”
Sol Medium: Competent research, drafting, summarization, and revisions.
Sol Light: Straightforward factual or editorial work.
Luna: Mechanical extraction and reformatting.
For something like “return these 20 CPT codes, one per line,” Luna or Sol Light is perfectly adequate. For “What is the real strategic path for this $300 TROP2 assay when PLA is blocked by physician work?” use Astra Medium or Extra High.
What to ignore
GPT-5.6 Sol, Terra, and Luna may remain visible temporarily while GPT-6 rolls out. They are the outgoing generation; there is little reason for you to choose them for a new project when the corresponding GPT-6 option is available. GPT-5.5 is scheduled to disappear from ChatGPT, Work, and Codex on October 14, 2026. (ChatGPT Learn)
One important distinction: GPT-6 Sol and Luna launched today in ChatGPT Work and Codex, not ordinary Chat mode. Availability is still rolling out, so what appears depends on the particular screen, account, and workspace. (ChatGPT Learn)
Bottom line: Choose Astra Medium and largely forget the picker exists. Escalate to Astra Extra High for the small pile of doom that needs to become a thoughtful, defensible essay.
