Appearance
OpenAI
Codex Plus Pool — 0.03x of official OpenAI pricing. Supports gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, gpt-5.5, gpt-5.4, and gpt-5.4-mini.
Codex Pro Pool — 0.05x of official OpenAI pricing. Supports all models above, plus gpt-5.3-codex-spark.
Plus Pool and Pro Pool use the same real OpenAI models as the official API. There is no difference in model quality. Pro Pool is a backup for when Plus Pool is unstable.
Image generation is available through a dedicated image pool at $0.015 per generated image via CE Image Studio or the /v1/images API.
Official OpenAI Pricing
Official OpenAI pricing for reference. Per 1M tokens, USD.
| Model | Context | Cache Read | Cache Write | Input | Output |
|---|---|---|---|---|---|
| gpt-5.6-sol* | 1M | $0.50 | $6.25 | $5.00 | $30.00 |
| gpt-5.6-terra | 1M | $0.20 | $2.50 | $2.00 | $12.00 |
| gpt-5.6-luna | 1M | $0.02 | $0.25 | $0.20 | $1.20 |
| gpt-5.5 | 272K | $0.50 | — | $5.00 | $30.00 |
| gpt-5.4 | 1M | $0.25 | — | $2.50 | $15.00 |
| gpt-5.4-mini | 272K | $0.075 | — | $0.75 | $4.50 |
| gpt-5.3-codex-spark | 128K | $0.175 | — | $1.75 | $14.00 |
CE Plus Pool Pricing (0.03x)
| Model | Cache Read | Cache Write | Input | Output |
|---|---|---|---|---|
| gpt-5.6-sol | $0.015 | — | $0.15 | $0.90 |
| gpt-5.6-terra | $0.006 | — | $0.06 | $0.36 |
| gpt-5.6-luna | $0.0006 | — | $0.006 | $0.036 |
| gpt-5.5 | $0.015 | — | $0.15 | $0.90 |
| gpt-5.4 | $0.0075 | — | $0.075 | $0.45 |
| gpt-5.4-mini | $0.00225 | — | $0.0225 | $0.135 |
CE Pro Pool Pricing (0.05x)
| Model | Cache Read | Cache Write | Input | Output |
|---|---|---|---|---|
| gpt-5.6-sol | $0.025 | — | $0.25 | $1.50 |
| gpt-5.6-terra | $0.01 | — | $0.10 | $0.60 |
| gpt-5.6-luna | $0.001 | — | $0.01 | $0.06 |
| gpt-5.5 | $0.025 | — | $0.25 | $1.50 |
| gpt-5.4 | $0.0125 | — | $0.125 | $0.75 |
| gpt-5.4-mini | $0.00375 | — | $0.0375 | $0.225 |
| gpt-5.3-codex-spark | $0.00525 | — | $0.0525 | $0.42 |
Notes
* OpenAI's promotional pricing on gpt-5.6-sol does not apply to Codex subscriptions, so CE is unable to pass on the discount. Prices shown above reflect subscription credit rates.
GPT-5.6 cache writes are 1.25x input with a guaranteed 30-minute cache lifetime. CE does not charge for cache writes.
Context over 272K tokens is billed at 1.5–2x by OpenAI. CE applies the same long context rates.
Based on our testing, GPT-5.6 series models show noticeable quality degradation above 400K context.
The Codex client defaults to a 258K context limit — using 1M context requires manual configuration.
Caching is optimized for Codex and OpenCode. Third-party clients such as Hermes may experience lower cache hit rates — please test on your own.
If calling the API directly, pass session_id and prompt_cache_key to enable cache hits. You can observe how the Codex client constructs these fields to guide your own implementation.
CE's GPT API is powered by Codex. If you do not specify the instruction field, Codex's default system prompt will be used. To use your own system prompt, set the instruction field explicitly.