For an initial GPT API shortlist, consider GPT-6 Astra (gpt-6-astra), GPT-6.1 Sol (gpt-6.1-sol), GPT-6.1 Terra (gpt-6.1-terra), and GPT-6 Luna (gpt-6-luna). The available descriptions position Astra for demanding analysis and ambitious deliverables, Sol for complex work, Terra for visual reasoning and computer use, and Luna for focused, high-volume tasks. This is a shortlist, not a complete list of every model or specialty offering. (OpenAI model-selection guide)
API availability is distinct from availability in ChatGPT or other OpenAI products. OpenAI notes that model availability, tools, reasoning settings, and usage limits can differ by product and model version; check the API model catalog for API availability. (OpenAI model-selection guide)
GPT API models to shortlist
| Model | API model ID | Stated fit | Standard text-token price* |
|---|---|---|---|
| GPT-6 Astra | gpt-6-astra |
Demanding analysis and ambitious deliverables. | $10 per 1M input tokens; $50 per 1M output tokens |
| GPT-6.1 Sol | gpt-6.1-sol |
Complex work; positioned as offering near-Astra performance at lower cost. | $2 per 1M input tokens; $10 per 1M output tokens |
| GPT-6.1 Terra | gpt-6.1-terra |
Visual reasoning, precise computer use, and clear, natural writing. | $1 per 1M input tokens; $5 per 1M output tokens |
| GPT-6 Luna | gpt-6-luna |
Focused, high-volume tasks; described as OpenAI’s most efficient model. | $0.10 per 1M input tokens; $0.50 per 1M output tokens |
The model names, IDs, and task descriptions are listed in OpenAI’s model documentation, Sol page, Terra page, and Luna page. The input and output rates appear in the API pricing documentation. OpenAI describes the prices on its API page as standard processing rates for context lengths under 272K. Rates are stated in dollars per million tokens, with input and output billed separately; these are API charges, not ChatGPT subscription prices. Check the pricing page for other processing modes and applicable conditions before estimating a workload. (OpenAI API pricing)
*Prices shown are standard text-token rates for the four models in this table, not every model or modality in the catalog. The all-models directory includes specialized offerings as well; don’t infer their IDs or prices from these entries. (OpenAI API pricing)
Which model should you shortlist?
- Start with Luna for focused, frequent, or high-volume tasks where efficiency matters. Treat this as a candidate, then check that its output meets your quality requirements. (GPT-6 Luna)
- Consider Terra for visual reasoning, computer-use workflows, writing based on supplied material, or scoped code fixes. For tasks that change records or pages, verify the saved result and confirm unrelated data was preserved. (GPT-6.1 Terra)
- Compare Sol with Astra for complex work when quality and cost both matter. OpenAI positions Sol as offering near-Astra performance at lower cost; that is provider positioning, not a guarantee of a particular saving or result on your workload. Compare them on the same task. (GPT-6.1 Sol)
- Consider Astra for demanding analysis or ambitious deliverables when capability is the priority and cost and latency are less important. Validate that choice against your task and budget. (OpenAI model-selection guide)
These are starting points, not universal rankings. OpenAI’s model-selection guide recommends considering workflow frequency, required speed, intended use of the output, and quality expectations. It also cautions that model availability and settings can vary by product and version. (OpenAI model-selection guide)
How to compare models for a production task
- Choose representative inputs. Include ordinary cases and difficult cases that matter to your application.
- Set acceptance criteria first. For writing, check facts, qualifications, audience, and voice. For computer use, inspect saved changes and confirm the completion report matches what happened. For code, run regression tests; for interactive interfaces, inspect the rendered result on desktop and mobile.
- Compare candidates on the same inputs. Use the same criteria for each model rather than relying on one impressive example.
- Measure the whole workflow. Consider output quality and completion time, including time spent reviewing or repairing results. Keep a candidate only if it meets your quality bar and practical constraints.
OpenAI recommends repeating comparisons across representative tasks and accounting for review and repair time as well as completion time. Its documentation does not prescribe a sample size or numerical pass threshold, so set those for your use case. (OpenAI model-selection guide)
Where to check the API model list and pricing
Use the official API model catalog to check model entries and IDs, the all-models directory to explore specialized categories, and OpenAI API pricing for applicable charges. Confirm a model’s API availability, exact ID, capabilities, and pricing before making it a production dependency or relying on a cost estimate.
Sources
GPT-6 Astra Model | OpenAI API
GPT-6.1 Sol Model | OpenAI API