Summary
OpenAI launched o3-mini on January 31, 2025. Users could choose low, medium, or high reasoning. The model could return set-format answers and use tools. At launch, the API cost $1.10 per million input tokens and $4.40 per million output tokens. An API lets software send requests to an AI service. A token is a piece of text the service reads or returns. This guide compares launch details with DeepSeek R1. Check the status note before choosing a model.
Current Status of o3-mini
As of October 19, 2026, OpenAI labels o3-mini as deprecated in its API model documentation. This means OpenAI marks it as an older option that developers may need to replace. The details below describe its January 2025 launch.
What made o3-mini stand out at launch?
OpenAI said o3-mini cost less than o1-mini. It could work through problems for software developers. Users could choose how much reasoning it used:
- Reasoning settings – Users could choose low, medium, or high. Higher settings gave the model more time to work through a problem.
- Structured outputs – Developers could request JSON fields for other tools to read.
- Function Calling – The model can ask outside tools to do tasks.
- Speed and tests – OpenAI said o3-mini ran 24% faster than o1-mini. It also scored higher on some math and coding tests.
- Free Access for ChatGPT Users – People on ChatGPT’s free plan could use this type of model for the first time.
DeepSeek R1 used a different design and price. Its launch also affected chipmakers’ market values.
DeepSeek R1 vs. OpenAI o3-mini at launch
Both were lower-cost options at launch. They differed in design, training, license, and price.
Comparison table: DeepSeek R1 vs. OpenAI o3-mini
| Feature | DeepSeek R1 | OpenAI o3-mini |
|---|---|---|
| Architecture | MoE; 37B of 671B parameters active per request | Smaller version of OpenAI’s o3 model |
| Training Approach | Reinforcement learning (feedback-based) | Deliberative alignment (safety-focused) |
| Performance | 79.8% on AIME 2024, 97.3% on MATH-500 | 71.7% on SWE-bench Verified, 96.7% on AIME |
| Input Context Window | 128,000 tokens | 200,000 tokens |
| Maximum Output Tokens | 32,000 tokens | 100,000 tokens |
| Cost per Million Input Tokens | $0.14 | $1.10 |
| Cost per Million Output Tokens | $0.55 | $4.40 |
| Usability | MIT license; users can change and share the model | ChatGPT free and paid plans |
| Best For | Research and analysis | Developers using JSON or tools |
In this table, B means billion. MoE means Mixture of Experts. Parameters are settings a model learns during training. A context window is the amount of text a model can use at once. Input is text sent to the AI; output is its answer. A token is a piece of text. A benchmark is a test. These results come from different tests, so they are not direct comparisons.
OpenAI vs. DeepSeek launch pricing
The prices below are the API rates reported at the models’ January 2025 launches. They are past comparisons, not a current price list. The charge is based on the small pieces of text sent to the AI and returned in its answer.
OpenAI priced o3-mini below o1-mini. DeepSeek R1’s listed API rates were lower still:
-
o3-mini Pricing:
- $1.10 per million input tokens
- $4.40 per million output tokens
-
DeepSeek R1 Pricing:
- $0.14 per million input tokens
- $0.55 per million output tokens
DeepSeek R1’s listed input and output rates were lower. OpenAI announced o3-mini at 63% below o1-mini’s price.
What the Launch Showed About OpenAI and DeepSeek
The launch-era comparison highlighted three differences:
- Pricing – o3-mini cost 63% less than o1-mini. DeepSeek R1’s listed API rates were lower.
- Test results – DeepSeek R1 led several public tests cited at launch. OpenAI said o3-mini was faster than o1-mini.
- Access – ChatGPT offered o3-mini on free and paid plans at launch.
What Business Owners Could Learn at Launch
At launch, o3-mini offered API developers several potential benefits:
- Lower Development Costs – OpenAI announced a lower price than its earlier o1-mini model.
- Math and coding tests – OpenAI said o3-mini outscored o1-mini on some math and coding tests.
- Free Plan Access – ChatGPT’s free plan added access to a model that could work through problems.
- Fixed-Format Answers and Tool Use – Developers could request JSON or connect outside tools. See our AI tools guide for more examples.
FAQ
How much does OpenAI o3-mini cost?
At launch, o3-mini cost $1.10 per million input tokens and $4.40 per million output tokens. That was 63% less than o1-mini. Input is text sent to AI; output is its answer.
What is o3-mini?
OpenAI launched o3-mini on January 31, 2025. It had low, medium, and high reasoning settings. It could return set-format answers and use tools. ChatGPT users could also use it on the free plan.
How does o3-mini compare to DeepSeek R1 on price?
In January 2025, DeepSeek listed input at $0.14 per million tokens. Output cost $0.55 per million tokens. Input is text sent to the AI; output is its answer.
What to Remember About o3-mini at Launch
At launch, o3-mini cost less than OpenAI’s o1-mini. DeepSeek R1 listed even lower API rates. These prices reflect January 2025. Check current rates before choosing an API. The o3 family also powers OpenAI’s Deep Research, which compiles reports on its own.
