Short answer: shortlist by workload
OpenAI’s current API model-selection guide positions GPT-6 Astra (gpt-6-astra) for the most demanding reasoning, coding, and professional work; GPT-6.1 Sol (gpt-6.1-sol) for complex work at lower cost; GPT-6.1 Terra (gpt-6.1-terra) for visual reasoning, precise computer use, and clear writing; and GPT-6 Luna (gpt-6-luna) for focused, high-volume tasks. These are OpenAI’s qualitative recommendations, not a guarantee of best fit for your application. Shortlist by workload, then compare finalists on representative tasks. (developers.openai.com)
Compare the listed models
| Model | OpenAI’s stated positioning | Model ID | Standard input price | Standard output price | Listed context window | Maximum output |
|---|---|---|---|---|---|---|
| GPT-6 Astra | Demanding reasoning, coding, and professional work | gpt-6-astra |
$10 / 1M tokens | $50 / 1M tokens | 1.05M tokens | 128K tokens |
| GPT-6.1 Sol | Complex work at lower cost; coding, computer use, and professional work | gpt-6.1-sol |
$2 / 1M tokens | $10 / 1M tokens | 1.05M tokens | 128K tokens |
| GPT-6.1 Terra | Visual reasoning, computer use, and clear writing | gpt-6.1-terra |
$1 / 1M tokens | $5 / 1M tokens | 1.05M tokens | 128K tokens |
| GPT-6 Luna | Focused, high-volume tasks | gpt-6-luna |
$0.10 / 1M tokens | $0.50 / 1M tokens | 1.05M tokens | 128K tokens |
OpenAI’s catalog labels these as prices per “Input MTok” and “Output MTok”; model pages specify prices per 1 million tokens. The prices are displayed with a dollar sign, but the documentation reviewed here does not spell out a currency code. Treat the figures as listed rates, not a workload-specific estimate, and confirm applicable rates before budgeting. (developers.openai.com)
Pricing can vary with request details. For GPT-6 Astra, prompts with more than 272K input tokens are priced at 2× the input and cache rates and 1.5× the output rate for the full request. Astra’s page also lists Batch and Flex at 50% of Standard rates and Fast mode at 2× the applicable rates. Tool-specific models may have a fee per tool call. Check the current pricing details for your model and request configuration. (developers.openai.com)
The model catalog lists Functions, Web search, File search, and Computer use for each of the four models. It also displays reasoning-level options, but their availability and behavior can depend on the model and endpoint. Shared listed limits or tools do not establish equal performance or compatibility. (developers.openai.com)
A practical selection workflow: choose a shortlist from the model’s stated focus, verify that its capabilities and endpoint fit your application, then compare finalists on representative inputs. Estimate input and output usage separately using the applicable current rates; do not treat qualitative positioning as a benchmark or cost estimate.
Check model and API fit before production
OpenAI’s catalog says models are available through the Responses API and Client SDKs. The model pages for GPT-6 Astra, GPT-6.1 Sol, and GPT-6.1 Terra list both Chat Completions and Responses. For Sol and Terra, the model pages specify that tool calling requires the Responses API. Confirm support for your exact model, endpoint, and needed features in that model’s documentation; endpoint support alone does not establish that every tool works on both endpoints. (developers.openai.com)
Before deployment, confirm the selected model’s current availability and lifecycle, the capabilities your product needs, and any material access, rate-limit, or regional constraints. The catalog’s statement that “latest OpenAI models” support text and image input, text output, multilingual capabilities, and vision does not identify which versions it means; verify those capabilities for your chosen model. (developers.openai.com)
When to look beyond the flagship shortlist
OpenAI’s catalog also groups specialized offerings under image generation and editing, realtime speech and translation, transcription, cyber, and life sciences. These are signposts rather than detailed comparisons; consult the relevant current model documentation if your workload centers on one of those tasks. (developers.openai.com)
Bottom line: Astra, Sol, Terra, and Luna form a current, documented shortlist for different workloads—not a universal ranking. Compare the listed rates and capabilities, then verify your chosen model’s endpoint fit, availability, and request-specific pricing before production use.
Sources
GPT-6 Astra Model | OpenAI API
GPT-6.1 Sol Model | OpenAI API