OpenAI-compatible API · in preparation
Simple API.
Plain truth.
Eight lower-cost GPT models, with prices and request-log boundaries stated in the open.
Models / standard rates
Eight models.
Prices upfront.
USD per 1M tokens. Input and output are billed separately.
| Model | Input | Output |
|---|---|---|
gpt-5.4-mini | $0.2500 | $1.5000 |
gpt-5.6-luna | $0.3333 | $2.0000 |
gpt-5.2 | $0.5833 | $4.6667 |
gpt-5.3-codex | $0.5833 | $4.6667 |
| Model | Input | Output |
|---|---|---|
gpt-5.4 | $0.8333 | $5.0000 |
gpt-5.6-terra | $0.8333 | $5.0000 |
gpt-5.5 | $1.6667 | $10.0000 |
gpt-5.6-sol | $1.6667 | $10.0000 |
Rates are rounded to four decimals. Cached input is billed at 10% of standard input.
Context above 272k on GPT-5.4, GPT-5.5 and GPT-5.6 profiles: input ×2; output ×1.5.
Priority tier: ×2, or ×2.5 on GPT-5.5. Account-specific rates take precedence when configured.
GPT-5.6 Luna, Terra and Sol are GorgeousGPT routing profiles, not official OpenAI model IDs.
Why so cheap / under the hood
Subscriptions.
Shared capacity.
At launch, requests will use shared ChatGPT subscription capacity instead of buying every token at public API rates. That is why the price is lower.
- API request
- Pool account
- Compatibility layer
- ChatGPT
- Response
Model aliases and request transforms may apply. Shared capacity can run out at peaks. Written upstream authorisation is required before launch.
Try the interface. Read the boundaries.
One account for balance, API keys, usage, and current rates. Production launch follows written upstream approval.