ModelsSupportEnterpriseBlog
500+ AI Model API, All In One API.Just In CometAPI
Models API
Developer
Quick StartDocumentationAPI Dashboard
Resources
AI ModelsBlogEnterpriseChangelogAbout
2025 CometAPI. All right reserved.Privacy PolicyTerms of Service
Home/Models/OpenAI/O3 Pro
O

O3 Pro

Input:$16/M
Output:$64/M
Context:200K
Max Output:100K
OpenAI o3‑pro is a “pro” variant of the o3 reasoning model engineered to think longer and deliver the most dependable responses by employing private chain‑of‑thought reinforcement learning and setting new state‑of‑the‑art benchmarks across domains like science, programming, and business—while autonomously integrating tools such as web search, file analysis, Python execution, and visual reasoning within API.
New
Commercial Use
Overview
Features
Pricing
API
Versions

Basic Information & Features

  • Model Class: o3-Pro is part of OpenAI’s “reasoning models,” designed to think step-by-step rather than generate immediate responses.
  • Availability: Accessible via ChatGPT Pro/Team interfaces and the OpenAI developer API as of June 10, 2025.
  • Access Tiers: Replaces the previous o1-Pro edition; Enterprise and Edu users onboard in the week following launch.

Technical Details

  • Architecture: Builds on the o3 backbone with an enhanced private chain of thought, enabling multi-step reasoning at inference.
  • Tokenization: Supports the same token schema as its predecessors—1 million input tokens ≈ 750,000 words.
  • Extended Capabilities: Includes web search, Python code execution, file analysis, and visual reasoning; image generation remains unsupported in this release.

Benchmark Performance

  • Math & Science: Surpassed Google Gemini 2.5 Pro on the AIME 2024 contest, demonstrating superior problem-solving in advanced mathematics.
  • PhD-Level Science: Outperformed Anthropic’s Claude 4 Opus on the GPQA Diamond benchmark, indicating robust expertise in scientific domains.
  • Enterprise Use: Internal tests report consistent wins over predecessor models across coding, STEM, and business reasoning tasks.

Technical Indicators

  • Latency: Response times are higher than o1-Pro—reflecting the deeper reasoning chains—averaging 1.5× the previous latency.
  • Throughput: Sustained token-generation throughput of up to 10 tokens/sec in burst mode.

With its enhanced reasoning chains, expanded feature set, and leading benchmark performance, o3-Pro represents a significant step forward in reliable, high-precision AI.

How to access o3-pro API

Step 1: Sign Up for API Key

Log in to cometapi.com. If you are not our user yet, please register first. Sign into your CometAPI console. Get the access credential API key of the interface. Click “Add Token” at the API token in the personal center, get the token key: sk-xxxxx and submit.

Step 2: Send Requests to o3-pro API

Select the “\**o3-pro\**” endpoint to send the API request and set the request body. The request method and request body are obtained from our website API doc. Our website also provides Apifox test for your convenience. Replace <YOUR_API_KEY> with your actual CometAPI key from your account.

Insert your question or request into the content field—this is what the model will respond to . Process the API response to get the generated answer.

Step 3: Retrieve and Verify Results

Process the API response to get the generated answer. After processing, the API responds with the task status and output data.

Features for O3 Pro

Explore the key features of O3 Pro, designed to enhance performance and usability. Discover how these capabilities can benefit your projects and improve user experience.

Pricing for O3 Pro

Explore competitive pricing for O3 Pro, designed to fit various budgets and usage needs. Our flexible plans ensure you only pay for what you use, making it easy to scale as your requirements grow. Discover how O3 Pro can enhance your projects while keeping costs manageable.
Comet Price (USD / M Tokens)Official Price (USD / M Tokens)Discount
Input:$16/M
Output:$64/M
Input:$20/M
Output:$80/M
-20%

Sample code and API for O3 Pro

The o3-Pro API is a RESTful ChatCompletion endpoint that enables developers to invoke OpenAI’s advanced chain-of-thought reasoning, code execution, and data-analysis capabilities via configurable parameters (model="o3-pro", messages, temperature, max_tokens, streaming, etc.) for seamless integration into complex workflows.

Versions of O3 Pro

The reason O3 Pro has multiple snapshots may include potential factors such as variations in output after updates requiring older snapshots for consistency, providing developers a transition period for adaptation and migration, and different snapshots corresponding to global or regional endpoints to optimize user experience. For detailed differences between versions, please refer to the official documentation.
version
o3-pro
o3-pro-2025-06-10

More Models

O

o4-mini-deep-research

O

o4-mini-deep-research

Input:$1.6/M
Output:$6.4/M
O4-Mini-Deep-Research is OpenAI’s latest agentic reasoning model, combining the lightweight o4-mini backbone with the advanced Deep Research framework. Designed to deliver fast, cost-efficient deep information synthesis, it enables developers and researchers to perform automated web searches, data analysis, and chain-of-thought reasoning within a single API call.
L

Llama-4-Scout

L

Llama-4-Scout

Input:$0.216/M
Output:$1.152/M
Llama-4-Scout is a general-purpose language model for assistant-style interaction and automation. It handles instruction following, reasoning, summarization, and transformation tasks, and can support light code-related assistance. Typical uses include chat orchestration, knowledge-augmented QA, and structured content generation. Technical highlights include compatibility with tool/function calling patterns, retrieval-augmented prompting, and schema-constrained outputs for integration into product workflows.
L

Llama-4-Maverick

L

Llama-4-Maverick

Input:$0.48/M
Output:$1.44/M
Llama-4-Maverick is a general-purpose language model for text understanding and generation. It supports conversational QA, summarization, structured drafting, and basic coding assistance, with options for structured outputs. Common applications include product assistants, knowledge retrieval front-ends, and workflow automation that require consistent formatting. Technical details such as parameter count, context window, modality, and tool or function calling vary by distribution; integrate according to the deployment’s documented capabilities.
M

Kimi-K2

M

Kimi-K2

Input:$0.48/M
Output:$1.92/M
- **kimi-k2-250905**: Moonshot AI's Kimi K2 series 0905 version, supporting ultra-long context (up to 256k tokens, frontend and Tool calls). - 🧠 Enhanced Tool Calling: 100% accuracy, seamless integration, suitable for complex tasks and integration optimization. - ⚡️ More Efficient Performance: TPS up to 60-100 (standard API), up to 600-100 in Turbo mode, providing faster response and improved Inference capabilities, knowledge cutoff up to mid-2025.
X

Grok 3 Reasoner

X

Grok 3 Reasoner

Input:$2.4/M
Output:$12/M
Grok-3 reasoning model, with chain-of-thought, Elon Musk's competitor to R1. This model supports a maximum context length of 100,000 tokens.
X

Grok 3 Mini

X

Grok 3 Mini

Input:$0.24/M
Output:$0.4/M
A lightweight model that thinks before responding. Fast, smart, and ideal for logic-based tasks that don't require deep domain knowledge. Raw thought traces are accessible. This model supports a maximum context length of 100,000 tokens.

Related Blog

New Veo3.1: More consistency ,diverse output and richer
Jan 14, 2026
veo-3-1

New Veo3.1: More consistency ,diverse output and richer

Google’s Veo 3.1 was updated in January, bringing focused improvements that push image-to-video workflows closer to production quality. The release emphasizes image-to-video fidelity, improved temporal and character consistency, native vertical output for mobile platforms, and higher-definition outputs via improved 1080p quality and a 4K upscaling path. For creators and developers who have been working around the “crop-then-edit” workflow for social vertical formats, Veo 3.1’s native 9:16 output and improved upscaling promise to reduce friction and deliver more polished, platform-ready clips.