Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.
Supported API features and tools
Streaming responses
Function calling and structured outputs
Text and image input
Web search, file search, image generation tools, and MCP
Before you choose
No native audio or video support; native output is text.
Code interpreter is not supported. Do not imply that the model executed a calculation or test without a separate execution tool.
Fine-tuning is not supported.
Difficult requests can take several minutes. The gpt-5-pro-2025-10-06 snapshot is scheduled to retire December 11, 2026.
Choose the reasoning effort
high · default
GPT-5 Pro defaults to high and supports only high reasoning effort. It is not a three-option Pro model like GPT-5.2 Pro. If your workflow needs a lower-effort tradeoff, compare another model rather than sending an unsupported setting.
Responses API only; Chat Completions is not supported. Batch is listed as supported, and background mode can help handle long-running requests.
Maximum output is 272,000 tokens, not GPT-5’s 128,000. Input, reasoning, and output still share the 400,000-token total context; maxima are not independent allowances.
The documented snapshot is gpt-5-pro-2025-10-06. API rate limits depend on usage tier; the free API tier is not supported.
Put it to work
Start with a more useful prompt.
Original examples from EZ Ai Assist. Adapt these to your task and the features available in your workspace.
Workflow 01
Review an architecture tradeoff
Make the decision conditional on explicit evidence and operating constraints.
Evaluate these architecture options against the supplied workload, reliability objectives, staffing constraints, and migration budget. Identify tradeoffs that could change under a different workload and cite the evidence behind each conclusion. Recommend one option only if the information supports it; otherwise identify the decisive missing measurements. End with a small validation plan and reversible next steps.
Workflow 02
Find a flaw in a formal specification
Search for counterexamples before implementation hides an inconsistency.
Inspect this formal specification for incompatible constraints, missing cases, and ambiguous state transitions. Construct a minimal example that violates each suspected invariant if possible. Cite the exact clauses involved and separate contradictions from underspecified behavior. Propose a narrowly scoped amendment and test cases without claiming to have executed a solver.
Workflow 03
Synthesize a technical review packet
Create a traceable analysis from multiple supplied engineering documents.
Synthesize these engineering review documents into a decision packet. Separate verified observations, competing explanations, and proposed actions. For each recommendation, cite its sources, state its dependencies, and describe what evidence would make it wrong. Preserve disagreements between documents instead of silently resolving them. Finish with the three highest-value follow-up checks.
Developer reference
OpenAI API pricing
These are OpenAI API reference prices, not EZ Ai Assist subscription prices.
These are Standard GPT-5 Pro rates, not GPT-5 or the newer GPT-5.6 Sol Pro mode.
The reviewed table lists no cached-input discount or separate long-context rate. An absent price is not a zero-cost allowance.
Reasoning contributes to output usage. Tool calls and other service tiers can change the bill; compare measured cost and quality when migrating.
Common questions
A few things worth knowing.
Does GPT-5 Pro offer multiple reasoning efforts?
No. The model reference says it defaults to and only supports high. Medium and xhigh belong to other Pro models’ option sets, not this one. Use a different model if you need a lower-effort configuration.
What is its maximum output?
The documented ceiling is 272,000 tokens, within a 400,000-token total context. This differs from GPT-5’s 128,000-token output limit. Input, reasoning, and visible output must fit the applicable limits; maximum values are not additive guarantees.
When should an existing integration migrate?
OpenAI announced retirement of gpt-5-pro-2025-10-06 on June 11, 2026, with shutdown scheduled for December 11, 2026. Test the recommended replacement, GPT-5.6 Sol with API reasoning.mode: pro, before that deadline.
Is Pro mode the same as an EZ Ai Assist plan?
No. GPT-5 Pro is a model name, and the replacement’s reasoning.mode: pro is a provider API setting. Neither establishes subscription entitlements or guarantees that EZ Ai Assist exposes the same controls.
Can it stream and return structured outputs?
Yes, both are supported according to this model reference. It uses Responses rather than Chat Completions; Batch is also listed. For requests that take minutes, consider an appropriate background-mode integration and validate output before taking action.
Can it run code to check its calculations?
Code interpreter is not supported. The model can discuss code and calculations, but you should not treat a written answer as proof of execution. Use an independently controlled execution environment and inspect the results when verification requires running code.
Why should I measure more than answer length?
Reasoning can consume output tokens before the final answer, and tool use can add charges. The Standard reference is $15 input and $120 output per million tokens. Measure actual usage and latency alongside correctness, separately from EZ Ai Assist subscription pricing.
Check the source
Official documentation
Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.