OpenAI models
Selected API model profiles with sources, verification dates, and explicit pricing assumptions.
API specifications do not determine model availability, limits, or billing in a chat interface, coding agent, or IDE. Check the actual access method.
| Model / API ID | Context / input limit | Max output | Input / output | Verified |
|---|---|---|---|---|
GPT-6 Astragpt-6-astra | 1.05M | 128K | $10 / $50 | Pricing source |
GPT-5.6 Solgpt-5.6-sol | 1.05M | 128K | $4 / $20 | Pricing source |
GPT-5.6 Terragpt-5.6-terra | 1.05M | 128K | $2 / $12 | Pricing source |
GPT-5.6 Lunagpt-5.6-luna | 1.05M | 128K | $0.2 / $1.2 | Pricing source |
Compare profiles on your task
Use the same task and acceptance checks. Compare confirmed defects, corrections, elapsed time, and usage. Provider names and similarly named reasoning levels are not a shared performance scale.
- GPT-6 Astra
Evaluate on demanding reasoning and coding tasks against your existing baseline.
Reasoning control: reasoning.effort (low, medium, high, xhigh, max).
Standard short-context text rates. Long-context, caching, regional processing and Fast mode are priced separately.
- API specifications do not establish the limits or availability of this model in ChatGPT, Codex, or an IDE host.
- GPT-5.6 Sol
Evaluate as a general development and reasoning profile.
Reasoning control: reasoning.effort (none, low, medium, high, xhigh, max).
Standard short-context text rates. Long-context, caching, regional processing and Fast mode are priced separately. Sol pricing is promotional through at least 2026-11-21.
- API specifications do not establish the limits or availability of this model in ChatGPT, Codex, or an IDE host.
- GPT-5.6 Terra
Evaluate when balancing task quality, latency, and cost.
Reasoning control: reasoning.effort (none, low, medium, high, xhigh, max).
Standard short-context text rates. Long-context, caching, regional processing and Fast mode are priced separately.
- API specifications do not establish the limits or availability of this model in ChatGPT, Codex, or an IDE host.
- GPT-5.6 Luna
Evaluate for bounded extraction and other high-volume tasks with strong checks.
Reasoning control: reasoning.effort (none, low, medium, high, xhigh, max).
Standard short-context text rates. Long-context, caching, regional processing and Fast mode are priced separately.
- API specifications do not establish the limits or availability of this model in ChatGPT, Codex, or an IDE host.