Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.
Supported API features and tools
Streaming, function calling, and structured outputs
Image input and prompt caching
Web search and file search
Code interpreter, hosted shell, and apply patch
Skills, MCP, and tool search
Computer use and image generation through tools
Before you choose
No native audio or video support.
Fine-tuning is not supported.
Tool support does not mean a tool is enabled in every application.
Set an escalation rule for ambiguous or consequential work rather than relying on low cost alone.
Choose the reasoning effort
none · defaultlowmediumhighxhigh
Begin with the documented none setting for simple tasks, then test a reasoning level on the cases that fail your evaluation. Measure how many outputs need correction, not just how quickly they arrive. Keep the same evaluation set when changing settings.
Unsupported settings: minimal, max.
Responses, Chat Completions, and Batch are supported; the built-in tool list refers to Responses API integrations.
272,000 tokens is the documented maximum input, within a 400,000-token total context window.
The documented snapshot is gpt-5.4-mini-2026-03-17. Account-tier rate limits still apply.
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 one function for edge cases
Keep a coding task small and give the model the surrounding contract.
Review this function against the contract and examples below. Identify boundary cases, incorrect assumptions, and missing validation within this function only. For each confirmed issue, show a minimal input that demonstrates it and suggest a focused test. Distinguish bugs from optional style changes. If the available code is insufficient, name the missing dependency instead of guessing its behavior.
Workflow 02
Draft a reply from approved facts
Use a fixed knowledge boundary for a repeatable support-writing task.
Draft a short customer reply using only the approved facts below. Answer the customer’s specific question in plain language, include the next supported action, and do not invent policies, dates, discounts, or account details. If the facts do not support an answer, return a brief escalation note and the missing information instead. Keep the reply under 120 words.
Workflow 03
Check a screenshot for a known issue
Ask a narrow visual question with a clear uncertainty fallback.
Inspect this screenshot for the checklist items I provide. Return one row per item with status visible-pass, visible-fail, or cannot-determine, plus a short observation and screen location. Do not infer hover states, keyboard behavior, or content outside the image. For every cannot-determine item, name the smallest interactive check that would settle it.
Developer reference
OpenAI API pricing
These are OpenAI API reference prices, not EZ Ai Assist subscription prices.
The reviewed Standard table lists one set of rates for Mini, without a separate long-context pricing column.
Cached-input discounts apply to eligible cached tokens, not every repeated instruction. Tool calls can have additional fees.
Regional processing adds 10% where available. Other service tiers have separate rates; confirm current pricing before use.
Common questions
A few things worth knowing.
What makes a good Mini workflow?
Choose tasks with a clear input, a bounded output, and an inexpensive way to verify the result. Keep examples of failures, measure correction rates, and define an escalation path before increasing volume.
Does Mini support computer use and tool search?
Yes, both appear in its Responses API tool list. They still require integration and permission. Nano has a different tool list, so the two guides should not be treated as interchangeable.
How much material can I send?
The model page lists a 272,000-token maximum input, 400,000-token context window, and 128,000-token maximum output. Total context and maximum input are distinct limits, not two names for the same budget.
Is xhigh reasoning available?
Yes. Mini supports none, low, medium, high, and xhigh, with none as the documented API default. Test whether higher effort improves your actual task enough to justify its cost and latency.
Can I use image inputs?
Yes, image input is supported and native output is text. A screenshot can provide visual evidence, but it cannot establish interactions such as keyboard focus or a successful form submission.
Are the listed rates included in my subscription?
These are direct OpenAI API rates, not EZ Ai Assist subscription pricing or included usage. Check the app and the EZ Ai Assist pricing page for current plan details.
How do I avoid plausible but unsupported replies?
Provide the allowed facts, specify an uncertainty outcome, and test questions that cannot be answered from those facts. Review the model’s citations or evidence against the original material before relying on it.
Check the source
Official documentation
Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.