Bottom line: OpenAI’s API documentation marks GPT-4.5 Preview as shut down. Its model ID, gpt-4.5-preview, is not a production API target. For a current candidate to evaluate, OpenAI documents GPT-6.1 Terra with the model ID gpt-6.1-terra and support for the Responses API. Test that candidate against your own workload before migrating; model specifications do not guarantee equivalent results. (developers.openai.com)
Can you use GPT-4.5 in OpenAI’s production API?
No. OpenAI’s model page marks GPT-4.5 Preview as shut down. The page lists its former endpoint support, but that historical information does not make the model callable now. OpenAI announced the planned API phaseout in April 2025; the current model-page status, rather than the old announcement, is the relevant signal for a production decision. (developers.openai.com)
API access and ChatGPT access are separate. Historical reporting about GPT-4.5 in ChatGPT does not establish that it can be called through the API, or confirm its current ChatGPT availability.
What GPT-4.5 was documented to do
OpenAI’s model page describes GPT-4.5 Preview as accepting text and image input and producing text, with a 128,000-token context window and up to 16,384 output tokens. These are specifications for a model that the same page marks as shut down—not a reason to build a new integration on it. (developers.openai.com)
Earlier reporting attributed strengths such as writing, persuasiveness, and natural interaction to GPT-4.5. Those dated, qualitative descriptions do not demonstrate how it would perform on your workload or establish that an alternative will match its outputs.
A documented API candidate to evaluate
OpenAI’s current model catalog lists GPT-6.1 Terra as a fast, efficient model for visual reasoning and writing. Its documented request model ID is gpt-6.1-terra. The model page lists text and image input, text output, a 1,050,000-token context window, and a maximum of 128,000 output tokens. It supports both Chat Completions and Responses; OpenAI specifies that tool calling requires the Responses API. (developers.openai.com)
The model page lists this request target, but account-specific access is not guaranteed by the listing. Confirm gpt-6.1-terra appears in your API project’s available models and check your project’s applicable rate limits before rollout. Then evaluate it on representative prompts, including difficult cases, and compare output quality, latency, and token use against your requirements. (developers.openai.com)
GPT-4.5 API pricing—and how to read older comparisons
OpenAI’s GPT-4.5 model page lists historical standard rates of $75 per million input tokens, $37.50 per million cached input tokens, and $150 per million output tokens. These are not a current budget estimate for GPT-4.5: the model is shut down. (developers.openai.com)
For clarity, OpenAI’s API pricing documentation—not the GPT-4.5 model page—lists GPT-4.1 at $2 per million input tokens, $0.50 per million cached input tokens, and $8 per million output tokens. GPT-4.1’s model page now marks it deprecated, so treat these rates as a documented comparison, not a suggestion to start a GPT-4.1 integration. (developers.openai.com)
For a workload using 1 million uncached input tokens and 200,000 output tokens, the listed rates produce this illustrative historical comparison:
- GPT-4.5:
(1 × $75) + (0.2 × $150) = $105 - GPT-4.1:
(1 × $2) + (0.2 × $8) = $3.60
For comparison, GPT-6.1 Terra’s listed standard rates are $1 per million input tokens and $5 per million output tokens—or $2 for the same illustrative token mix. OpenAI notes that requests with more than 272,000 input tokens have different pricing multipliers, so this simple estimate does not cover long-context requests, caching, or other applicable charges. Check the current pricing page for the model and processing option you actually select. (developers.openai.com)
Production decision checklist
- Use a currently listed model ID. Do not use
gpt-4.5-previewfor a new API integration; evaluategpt-6.1-terraif its documented modalities and capabilities suit the task. - Confirm access and the endpoint. Check the model in your API project. Use the Responses API if your integration requires tool calling; otherwise confirm the endpoint you plan to use in the model documentation.
- Test representative work. Compare outputs on routine and difficult examples using explicit quality criteria, and measure latency and token use.
- Estimate cost from current rates. Use your measured input/output volumes and account for caching, long-context pricing, and any other applicable charges.
- Migrate only when results meet your requirements. A documented model target is a starting point, not a guarantee of compatibility or equal performance.
Decision: GPT-4.5 is not an option for a new OpenAI API deployment because its model page marks it shut down. Evaluate gpt-6.1-terra as a documented candidate, verify project access, and base any migration on current specifications, pricing, and workload tests. (developers.openai.com)
Sources
GPT-4.5 Preview Model | OpenAI API