o3-mini is a compact reasoning model from OpenAI designed for technical tasks, with a focus on mathematics, science, code, and multi-step logic problems. It offers adjustable reasoning effort and supports function calling, structured outputs, and streaming responses, making it suitable for embedding in coding assistants, problem-solving applications, and technical analysis workflows. Applications can integrate using the public request format in this page's API section.
Point the Base URL to api.acedata.cloud, configure your platform API key and the model ID below, and use your compatible SDK or client.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["ACEDATACLOUD_API_KEY"],
base_url="https://api.acedata.cloud/v1",
)
response = client.responses.create(
model="o3-mini",
input="Hello!",
)
print(response.output_text)Choose an available protocol for this model. OpenAI SDK uses a Base URL ending in /v1; Anthropic SDK uses the root URL. See each guide for protocol-specific parameters, tools and response formats.
Clarify capacity, inputs and outputs, and invocation methods before choosing a model.
The native capabilities on this page apply to o3-mini. Chat Completions uses messages, while Responses uses input; applications provide history according to the respective protocol, and shared parameters do not mean this model supports every combination.
Learn what o3-mini can bring to your work.
o3-mini focuses on mathematical, scientific, and programming reasoning, rather than merely generating fluent text. For problems involving multiple conditions, you can ask it to organize solution steps, explain assumptions, and provide verification methods. Clearly stating variables, units, and constraints makes it easier to obtain verifiable technical answers.
low, medium, and high allow the same model to adapt to tasks of different complexity. For routine code explanations, start with low; for analysis involving multiple constraints, start with medium; and try high for difficult problems. Increasing the level is a choice to invest more in reasoning; it does not mean you will get better answers every time, nor does it mean longer responses are required.
Function calling and structured outputs make o3-mini suitable for handling analysis stages within applications. You can define function names and parameters for the model to generate call suggestions; you can also specify a JSON structure to receive classifications, repair suggestions, or problem-solving results. Tool execution and result validation remain the application's responsibility; generating call parameters does not automatically complete an operation.
Start with specific tasks to find where the model can be effective.
Provide relevant code, error logs, expected behavior, and runtime conditions, and let o3-mini analyze possible causes and produce modification plans and test case suggestions. It is suitable for identifying algorithmic boundaries, missing conditions, and implementation logic issues. Fixes that require actual execution should be verified in a test environment rather than deploying generated code directly to production.
Submit the problem statement, formulas, known conditions, and your own attempted solution process, and let the model provide solution explanations, compare different methods, or check for flaws in derivations. Deliverables can include step-by-step explanations, assumption lists, and verification approaches. If a problem comes from an image, convert it to text first and retain necessary descriptions of symbols, units, and geometric relationships.
Organize business rules, configuration conditions, or algorithm requirements into text, and let o3-mini analyze relationships between conditions to generate decision tables, explanations of exception branches, or structured validation results. It is suitable for technical analysis with clearly defined inputs and acceptance criteria; when using JSON output, define fields and allowed values first, then perform programmatic validation.
Choose based on task complexity, input materials, and expected results.
If you previously used o1-mini for code, math, or science problems, o3-mini is worth considering as a replacement. OpenAI's published evaluations show improved answer accuracy and clarity in these areas, along with added support for function calling, structured outputs, and developer messages. It is recommended to validate performance using your own problem sets, code samples, and formatting requirements before deciding the scope of migration.
If you focus on technical problems and want flexible control over reasoning effort, choose o3-mini; for broader general-knowledge reasoning, consider o1. In some technical evaluations, o3-mini's medium setting performs close to o1, but this should not be taken to mean they are equivalent for all tasks. o3-mini-high is the high-effort option in ChatGPT; here, use o3-mini with high rather than switching to that name.
Based on o3-mini's characteristics, first validate a small task whose results can be checked.
You can ask directly: Explain this algorithm's behavior with empty input, duplicate elements, and extreme values, and provide time and space complexity plus the smallest counterexample. Use only the provided text and code.
Do not rely on image input; provide failure cases and output constraints, and choose reasoning effort based on task difficulty.
Use the full model ID o3-mini, first confirm the public request format and available parameters on the API page, then connect your application. Keep the result parsing, exception handling, and relevant evidence, and evaluate with the same set of real samples whether it is suitable for continued use.
Before formal use, understand the output quality and capability scope.
Answers to common questions about using o3-mini.
Start with medium and compare results on real tasks; try low for simple explanations, and high for complex derivations or difficult coding problems. Chat Completions uses reasoning_effort, while Responses uses reasoning configuration settings. o3-mini's native levels are low, medium, and high; other levels should not be applied to it.
It cannot directly understand screenshots because it does not support vision. First convert the code and errors in the screenshot to text, then submit file snippets, the runtime environment, and expected results. If the screenshot involves layout, graphics, or other information that cannot be fully transcribed, use a model that supports image understanding instead.
Yes, it supports structured outputs. When using Chat Completions, you can select json_schema through response_format and define the field structure, which is suitable for returning diagnostic results, rule classifications, or repair suggestions. Meeting the required structure does not guarantee that the content is correct; applications should still check field values and business conditions.
Use Chat Completions or Responses and provide the complete model ID. Chat Completions uses messages and choices, while Responses uses input and its corresponding response structure; handle history management, streaming events, and tool parameters separately according to the selected API, and do not mix the two formats.
o3-mini-high is the high reasoning-effort option provided in ChatGPT when OpenAI released it, not a model ID that needs to be substituted here. When calling it, still select o3-mini and set reasoning effort to high. It is also different from o3, and their names or capabilities should not be treated as exactly the same.
Choose a development approach or application suited to this model and follow its setup guide.
↗OpenAI Chat Completions API
SDK installation, authentication and examples↗OpenAI Responses API
SDK installation, authentication and examples
↗Using Ace Data Cloud with Cherry Studio
Chat client setup · Base URL, API key and model
↗Using Ace Data Cloud with ChatBox
Chat client setup · Base URL, API key and modelThis is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs