As of November 11, 2026: OpenAI lists gpt-5 as deprecated and gpt-5-chat-latest as shut down. For a new integration, check the current model guidance and select a supported model rather than relying on either historical ID.
What is the difference?
gpt-5 is OpenAI's earlier reasoning model for coding and agentic work. Its API documentation lists configurable reasoning effort: minimal, low, medium, and high.
gpt-5-chat-latest pointed to the GPT-5 snapshot previously used in ChatGPT. It was an API model ID, not a subscription plan or a setting that enables the ChatGPT interface. OpenAI now marks it as shut down.
| API model | Documented status | Context window | Maximum output | Input / output |
|---|---|---|---|---|
gpt-5 |
Deprecated | 400,000 tokens | 128,000 tokens | Text and images / text |
gpt-5-chat-latest |
Shut down | 128,000 tokens | 16,384 tokens | Text and images / text |
The Chat model page also lists a 272,000-token maximum input, which exceeds its stated 128,000-token context window. Because the documentation does not explain that inconsistency, do not use the input figure for capacity planning without confirming it with OpenAI.
Sources: GPT-5 model page and GPT-5 Chat model page.
How did API pricing compare?
Both model pages list the same text-token rates, per million tokens:
| Token type | Listed rate |
|---|---|
| Input | $1.25 |
| Cached input | $0.125 |
| Output | $10.00 |
These are API token rates, not ChatGPT subscription prices. Tool calls and other services can have separate charges; check the OpenAI pricing page for the current billing terms. The shut-down model's rates are historical information, not an assurance that it can still be used.
Which should you choose?
For a new production application, start with the current model guidance. The GPT-5 page recommends GPT-6 Astra; verify its current availability, pricing, features, and suitability for your workload before migrating. For an existing integration, locate calls to either legacy model ID, check provider deprecation notices, and test a supported replacement against representative inputs, quality targets, latency, and cost.
If you are comparing historical results, compare outputs from the exact model IDs and prompts you recorded. The model name alone does not establish that a replacement behaves identically.
FAQ
Is GPT-5 Chat the same as ChatGPT?
No. The documented gpt-5-chat-latest API ID referred to a GPT-5 snapshot previously used in ChatGPT. ChatGPT is a product, while an API model ID identifies a model used by an application.
Are the two models available now?
As of the date above, the model pages label gpt-5 deprecated and gpt-5-chat-latest shut down. Availability can change, and access may vary by account; consult the current official pages before deployment.
Editorial status: manually prepared; pending human review. Not published. Pricing, model status, and limits must be checked again before publication.
Sources
https://developers.openai.com/api/docs/models
https://developers.openai.com/api/docs/models/gpt-5
https://developers.openai.com/api/docs/models/gpt-5-chat-latest