OpenAI previous Pro model
GPT-5.2 Pro
A higher-compute GPT-5.2 variant for difficult professional problems that benefit from deeper reasoning, multi-turn analysis, and more precise answers.
- Context window
- 400K
- tokens
- Max output
- 128K
- tokens
- Input
- $21.00
- per 1M tokens
- Cached input
- Not listed
- no published rate
- Output
- $168.00
- per 1M tokens
01 / Overview
What GPT-5.2 Pro Is
GPT-5.2 Pro is OpenAI's previous Pro model for complex professional work, designed to spend more compute on tough problems and produce smarter, more precise responses than the standard GPT-5.2 tier.
A Higher-Compute Layer Above Standard GPT-5.2
The central difference between GPT-5.2 and GPT-5.2 Pro is not a larger context window or a broader modality set. It is the amount of inference effort the Pro variant is designed to spend on difficult requests.
OpenAI describes GPT-5.2 Pro as a model for problems that benefit from harder thinking. That makes it most relevant when the cost of an incorrect or shallow answer is higher than the cost of slower, more expensive inference.
The model is available through the Responses API and was designed to support multi-turn model interaction before returning a final answer. OpenAI warns that some requests can take several minutes and recommends background mode to avoid timeouts.
GPT-5.2 Pro is now a previous Pro generation rather than OpenAI's latest premium reasoning model. The current model card recommends GPT-5.5 Pro for developers who want the newer Pro tier.
- Higher-compute GPT-5.2 variant for difficult professional problems.
- Designed for more precise answers rather than minimum latency.
- Supports medium, high, and xhigh reasoning effort.
- Available through the Responses API.
- Some requests may take several minutes to complete.
- Provider
- OpenAI
- Family
- GPT-5.2
- Tier
- Pro
- Positioning
- Previous Pro model
- Knowledge cutoff
- Aug 31, 2025
- Input modalities
- Text, Image
- Output modality
- Text
02 / Use cases
Where GPT-5.2 Pro Fits Best
GPT-5.2 Pro is best reserved for problems where deeper reasoning has measurable value: difficult analysis, high-stakes technical decisions, complex synthesis, and escalation cases that standard models solve inconsistently.
Use Pro Selectively, Not as the Default for Every Request
A premium reasoning model is most valuable when it sits behind a routing or escalation policy.
Routine requests can often be handled by a cheaper model. GPT-5.2 Pro becomes relevant when the application identifies a task with unusual ambiguity, many interacting constraints, high business value, or a greater cost of failure.
Complex technical investigations are a strong example. A request may contain architecture constraints, logs, source fragments, previous attempts, and competing hypotheses. The model must weigh evidence rather than simply retrieve or transform information.
Decision-heavy professional analysis is another fit. The model can be evaluated on tasks where the answer must reconcile several sources, preserve assumptions, compare alternatives, and produce a precise conclusion.
Long-running research and synthesis also align with the model's Responses API design, especially when the system can run the work asynchronously rather than forcing an interactive timeout.
- Complex technical investigations with several competing explanations.
- High-value professional analysis with many constraints.
- Long-form synthesis across substantial source material.
- Escalation cases where cheaper models fail quality thresholds.
- Difficult decisions where precision matters more than response speed.
- 01
Technical investigation
Analyze interacting evidence, constraints, failure modes, and competing hypotheses before reaching a conclusion.
- 02
Decision analysis
Compare alternatives and preserve assumptions when the cost of a weak answer is materially higher than inference cost.
- 03
Complex synthesis
Reconcile multiple documents or evidence sources into a precise, internally consistent professional response.
- 04
Escalation tier
Route the hardest cases to Pro after a lower-cost model, evaluator, or business rule identifies insufficient quality.
03 / Pricing
GPT-5.2 Pro Pricing
GPT-5.2 Pro is priced at $21.00 per million input tokens and $168.00 per million output tokens.
Evaluate Cost per Successful Outcome
GPT-5.2 Pro is substantially more expensive than standard GPT-5.2, so token price alone is not enough to decide whether it belongs in a production workflow.
The better question is whether additional reasoning reduces retries, improves difficult decisions, avoids human escalation, or solves a class of requests that a cheaper model handles unreliably.
For example, a request with 10,000 input tokens and 2,000 output tokens costs about $0.546 before any additional service or tool charges.
The official model card does not publish a separate cached-input price for GPT-5.2 Pro, so production cost models should not assume the discounted cached-input rate available on standard GPT-5.2.
- $21.00 per 1M input tokens.
- $168.00 per 1M output tokens.
- No separate cached-input rate is published on the model card.
- Long reasoning can increase total output-token usage.
- Measure cost per completed high-value task rather than cost per isolated request.
1M tokens · USD
- Input
- $21.00
- Cached input
- Not listed
- Output
- $168.00
Example: 10K input + 2K output
- Input cost
- $0.2100
- Output cost
- $0.3360
- Estimated total
- $0.5460
04 / Context
A 400K Context Window for Difficult Professional Problems
GPT-5.2 Pro provides a 400,000-token context window with up to 128,000 output tokens, allowing difficult reasoning tasks to keep substantial evidence and working state in a single request.
Large Context Matters When the Relationships Matter
The value of a 400K context window is not simply the amount of text that fits inside it.
For a difficult technical investigation, context may contain architecture notes, source fragments, logs, issue history, prior conclusions, and constraints. For a professional analysis, it may contain reports, policies, tables, assumptions, and competing recommendations.
GPT-5.2 Pro can keep a substantial amount of this evidence available while reasoning through the task.
Still, more context can increase cost and make irrelevant information harder to separate from the material that actually matters. Retrieval, preprocessing, and deliberate context selection remain useful even when the entire source set technically fits.
- 400,000-token context window.
- Maximum output of 128,000 tokens.
- Suitable for large evidence sets and complex professional state.
- Useful for multi-document and multi-source reasoning.
- Context selection remains important for cost and signal quality.
Context window
400,000
Max output
128,000
A Pro workload can combine source documents, technical evidence, assumptions, prior analysis, requirements, and other context required to resolve a difficult problem.
05 / Reasoning
Medium, High, and XHigh Reasoning
GPT-5.2 Pro supports three reasoning-effort levels: medium, high, and xhigh.
Every Available Setting Is a Deliberate Reasoning Tier
Unlike standard GPT-5.2, the Pro variant does not expose none or low reasoning.
That reflects its intended role. GPT-5.2 Pro is not designed to become a low-cost or low-latency model for routine work. Even the lowest available setting begins at medium.
medium is a useful baseline for evaluating whether the Pro tier improves the hard cases in a workload. high and xhigh can then be tested on the subset of tasks where additional inference effort produces measurable gains.
The official model card does not specify a default reasoning level, so applications should set and record reasoning effort explicitly when running comparative evaluations.
Pro baseline
Start with medium when determining whether GPT-5.2 Pro materially improves the difficult cases that justify premium inference.
Maximum deliberation
Test high or xhigh for the most complex analytical and technical tasks when additional latency and token usage are acceptable.
06 / Capabilities
Responses API Features for Long-Running Reasoning
GPT-5.2 Pro supports text and image input, text output, streaming, function calling, and reasoning tokens through the Responses API, but it does not support Structured Outputs, fine-tuning, or predicted outputs.
Built Around Multi-Turn Reasoning Before the Final Response
OpenAI makes GPT-5.2 Pro available through the Responses API to support multi-turn model interactions before the model returns the final answer.
That design fits workloads where the model may need more than one reasoning step to work through a difficult problem.
Streaming is supported, which allows applications to surface output progressively when appropriate. Function calling is also supported, so the model can participate in application-controlled workflows that expose specific actions.
There are important differences from standard GPT-5.2. Structured Outputs are not supported on GPT-5.2 Pro. Fine-tuning and predicted outputs are also unavailable.
The model accepts image input in addition to text, making visual evidence available to reasoning workflows, but direct audio and video modalities are not supported.
- Responses API support.
- Text and image input with text output.
- Streaming supported.
- Function calling supported.
- Reasoning tokens supported.
- Structured Outputs are not supported.
- Fine-tuning is not supported.
- Predicted outputs are not supported.
- Supported
Multi-turn Responses
Use the Responses API for workflows where the model can reason across multiple interactions before returning the final answer.
- Supported
Streaming
Return model output progressively when the surrounding application benefits from streamed responses.
- Supported
Function calling
Connect model decisions to application-defined actions through supported function calls.
- Supported
Image input
Include visual evidence alongside text in difficult professional reasoning tasks.
- Not listed
Structured Outputs
OpenAI currently lists Structured Outputs as unsupported for GPT-5.2 Pro.
- Not listed
Fine-tuning
OpenAI currently lists fine-tuning as unsupported for GPT-5.2 Pro.
07 / Evaluation
Strengths and Limitations
GPT-5.2 Pro is a specialized premium reasoning tier: it can add value on difficult professional problems, but its high token price, longer execution time, Responses-only API surface, and newer Pro successors make selective routing essential.
Strengths
Higher-compute reasoning
GPT-5.2 Pro is explicitly designed for tough problems that benefit from harder thinking and more precise responses.
Three deep reasoning levels
Medium, high, and xhigh effort let teams evaluate how additional inference changes quality on difficult tasks.
Large professional context
A 400K-token window can hold substantial technical, documentary, and analytical evidence in one working context.
Multi-turn Responses design
The model is available through the Responses API to support multi-turn interactions before producing the final response.
What to consider
Premium token cost
At $21 input and $168 output per million tokens, GPT-5.2 Pro should be reserved for tasks where deeper reasoning has measurable value.
Long-running requests
OpenAI warns that some requests can take several minutes and recommends background mode to avoid timeouts.
No Structured Outputs
Unlike standard GPT-5.2, the Pro model does not support Structured Outputs.
Previous Pro generation
OpenAI now recommends GPT-5.5 Pro as the newer Pro model for developers selecting a current premium reasoning tier.
Evaluate the hard cases
Test GPT-5.2 Pro where additional reasoning can change the outcome
Run your most difficult analytical, technical, long-context, and decision-heavy prompts, then compare response quality, latency, token usage, and cost against newer Pro models in EidoStack.
Start FreeGPT-5.2 Pro remains documented as a previous Pro model. OpenAI recommends GPT-5.5 Pro for the latest Pro tier.
Common Questions
What is GPT-5.2 Pro?
GPT-5.2 Pro is OpenAI's previous Pro model for complex professional work. It uses additional compute to tackle tough problems and aims to produce smarter, more precise responses than standard GPT-5.2.
Is GPT-5.2 Pro deprecated?
The current OpenAI model card describes GPT-5.2 Pro as a previous Pro model but does not mark it as deprecated. OpenAI recommends GPT-5.5 Pro as the latest Pro model.
How much does GPT-5.2 Pro cost?
OpenAI lists $21.00 per 1M input tokens and $168.00 per 1M output tokens. The model card does not publish a separate cached-input price.
What is the context window of GPT-5.2 Pro?
GPT-5.2 Pro has a 400,000-token context window and supports up to 128,000 output tokens.
What reasoning levels does GPT-5.2 Pro support?
GPT-5.2 Pro supports medium, high, and xhigh reasoning effort.
What is the default reasoning level for GPT-5.2 Pro?
The current OpenAI model card lists medium, high, and xhigh as supported reasoning levels but does not specify a default.
What is the knowledge cutoff for GPT-5.2 Pro?
OpenAI lists August 31, 2025 as the knowledge cutoff for GPT-5.2 Pro.
Why can GPT-5.2 Pro requests take several minutes?
GPT-5.2 Pro is designed to spend more compute on difficult problems. OpenAI notes that some requests may take several minutes and recommends background mode to avoid timeouts.
Does GPT-5.2 Pro support image input?
Yes. GPT-5.2 Pro accepts text and image input and generates text output. Direct audio and video modalities are not supported.
Does GPT-5.2 Pro support streaming?
Yes. OpenAI lists streaming as supported.
Does GPT-5.2 Pro support function calling?
Yes. OpenAI lists function calling as supported.
Does GPT-5.2 Pro support Structured Outputs?
No. OpenAI currently lists Structured Outputs as unsupported for GPT-5.2 Pro.
Can GPT-5.2 Pro be fine-tuned?
No. OpenAI lists fine-tuning as unsupported for GPT-5.2 Pro.
What model does OpenAI recommend instead of GPT-5.2 Pro for the latest Pro tier?
The current GPT-5.2 Pro model card recommends GPT-5.5 Pro as OpenAI's newer Pro model.
Model information
Last updated
The specifications, pricing, reasoning levels, API behavior, and feature support on this page are based on OpenAI's official GPT-5.2 Pro model card and GPT-5.2 guidance. Provider pricing, availability, limits, and recommended replacement models may change, so production assumptions should be checked against the latest OpenAI documentation.