Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.
Supported API features and tools
Streaming responses and prompt caching
Function calling and structured outputs
Web search and file search
Code interpreter and hosted shell
Computer use and image generation tools
MCP, tool search, skills, and apply patch
Before you choose
No native audio or video support.
Image generation requires a tool; native output is text.
Fine-tuning is not supported.
Choose the reasoning effort
nonelowmedium · defaulthighxhigh
Start with the medium default and compare against a lower setting for straightforward tasks. Evaluate high or xhigh when the problem involves several dependent steps, using examples with known outcomes or a review rubric. Measure correctness, latency, and token use together. Specify the desired result and constraints rather than requesting a long answer as a proxy for careful work.
Unsupported settings: minimal, max.
Use the Responses API for built-in tools and multi-turn reasoning workflows. Chat Completions and Batch are also supported.
The documented default snapshot is gpt-5.5-2026-04-23. Pinning an available snapshot can make API evaluations easier to reproduce.
GPT-5.5 and GPT-5.5 Pro are separate models. Do not apply Pro's pricing or feature assumptions to this page. API rate limits depend on your usage tier.
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
Trace a cross-service failure
Connect logs, interfaces, and timing without turning missing evidence into a conclusion.
Investigate this failure across the supplied services using the logs, interface definitions, and recent changes below. Reconstruct the request path and timeline, then identify where observed behavior first differs from the expected contract. Cite evidence for each finding and separate confirmed causes from hypotheses. Propose the smallest diagnostic check for each unresolved hypothesis, followed by a minimal fix and regression plan if the cause is established. Do not invent missing logs, edit code, or change production systems.
Workflow 02
Stress-test a project decision
Evaluate a proposed direction against evidence, constraints, and the strongest alternative.
Review the proposed project decision using the evidence pack and constraints below. State the decision and success criteria, compare the proposal with its strongest practical alternative, and identify assumptions that could reverse the recommendation. Cite the source and section for every material factual claim. Distinguish verified evidence from estimates and unresolved disagreement. End with a recommendation, the cost of being wrong, and the smallest test that would reduce the most important uncertainty. Do not invent figures or make commitments.
Workflow 03
Plan a staged migration
Turn current-state constraints into small, reversible steps with explicit acceptance checks.
Create a staged migration plan from the current architecture to the target described below. Map dependencies, compatibility requirements, and data integrity risks. For each stage, define the smallest deliverable, acceptance tests, monitoring signals, and rollback conditions. Identify decisions requiring an owner and any missing information that blocks a safe sequence. Preserve existing behavior unless a change is explicitly requested. Finish with the first reversible step. Planning only: do not run commands, modify files, or deploy.
Developer reference
OpenAI API pricing
These are OpenAI API reference prices, not EZ Ai Assist subscription prices.
Above 272,000 input tokens, the increased rates apply to the full session, not just the excess tokens. OpenAI documents this rule for Standard, Batch, and Flex.
There is no additional cache-write charge for GPT-5.5; a separate cache-write rate is therefore not shown.
Regional processing adds 10%. This table covers Standard processing; other service tiers and tool usage can change the bill. Confirm current rates before estimating API spend.
Common questions
A few things worth knowing.
When is GPT-5.5 worth evaluating?
OpenAI positions GPT-5.5 for complex professional work and coding. Test it on tasks that expose the dependencies and ambiguity in your real workload, not just an easy demonstration. Compare the quality of the final deliverable, review effort, latency, and total cost. A newer or more expensive model is not automatically the best fit for every job.
How can I make a complex request easier to evaluate?
Lead with the outcome, then supply constraints, labeled evidence, and acceptance criteria. Ask the model to separate findings from assumptions and to surface missing information. Keep the final format reviewable: a short recommendation with supporting references is often easier to check than an undifferentiated narrative.
Does the large context window remove the need to organize inputs?
No. Prioritize relevant files and passages, label versions and dates, and identify which sources are authoritative. Request citations and check important claims against the originals. Keep enough room for the output and watch the long-context pricing threshold. Capacity for more material is not a guarantee of complete recall.
How does GPT-5.5 prompt caching differ from GPT-5.6?
GPT-5.5 has no additional cache-write charge and does not support explicit cache breakpoints. Its supported prompt_cache_retention value is 24h; documented cache lifetimes are typically around 30 minutes and can extend to 24 hours. A cache hit is not guaranteed. Use the current prompt-caching documentation when configuring direct API requests rather than copying GPT-5.6 cache controls.
Can GPT-5.5 review images and use external tools?
Image input and the listed tools are supported by the API, but the integration must make them available. Provide clear images and task context, and verify visual interpretations against the source. Require review before consequential external actions. A claim that a tool was used should be backed by an actual tool result, not merely a plausible answer.
How should I validate a proposed code or migration plan?
Check the affected interfaces, failure paths, and rollback assumptions against your actual system. Run focused tests in an appropriate environment before considering deployment. Ask for a distinction between checks already performed and checks only recommended. The example prompts on this page deliberately request review or planning, not permission to change production.
Does the API price table describe my EZ Ai Assist plan?
No. Direct OpenAI API billing is separate from an EZ Ai Assist subscription. The table is not a per-message quote and does not establish which tools, snapshots, or controls the app exposes. Consult our pricing page for subscriptions and the app for current availability. Share only source material you are authorized to provide.
Check the source
Official documentation
Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.