OpenAI Announces GPT-5 Release Within Months - Accelerating Reasoning Revolution with GPT-4.1 and o3/o4-mini
On April 4, 2025, OpenAI CEO Sam Altman announced a strategic plan change, revealing that while GPT-5’s release would be delayed by several months, o3 and o4-mini models would be released first1. This decision led to o3 and o4-mini becoming generally available on April 16, achieving breakthrough advances in AI reasoning capabilities.
Shift in Release Strategy and Rationale
CEO Altman announced, “Change of plans: we’ll release o3 and o4-mini first in a few weeks, then GPT-5 in a few months”1. Behind this strategic shift was the realization that GPT-5’s quality could be improved far beyond initial expectations, and the technical difficulty of integrating all features was greater than anticipated.
GPT-5 is designed as a complete multimodal system, eliminating the need to switch between reasoning models, standard language models, and image generation models like current ChatGPT, and is being developed to handle all functions in a single model.
Innovative Performance of the o3 Model
o3 is positioned as OpenAI’s most advanced reasoning model to date, becoming the first model capable of directly handling web browsing, image generation, and visual understanding2. Its performance in mathematics is particularly remarkable, achieving an 88.9% accuracy rate on the American Invitational Mathematics Examination (AIME) 2025.
The model also recorded an 83.3% score on the GPQA Diamond benchmark, which deals with PhD-level science problems, demonstrating advanced academic reasoning capabilities previously considered difficult for AI to achieve. These achievements have opened new possibilities for AI utilization in research fields.
Efficiency and Practicality of o4-mini
The o4-mini model was developed to provide performance exceeding o3 more efficiently. It achieved a 92.7% accuracy rate in mathematics tests, surpassing o3, and realized 81.4% high accuracy in science fields2.
From a cost-efficiency perspective, o4-mini is affordably priced at $1.10 per million input tokens and $4.40 per million output tokens, making it suitable for applications requiring large-scale processing.
Implementation of Agent Capabilities
o3 and o4-mini, as OpenAI’s first tool-enabled reasoning models, implement an agent-based AI approach. These models have the ability to directly execute web browsing, Python code execution, file operations, and image generation.
This capability enables breaking down complex tasks into steps and combining necessary tools for problem-solving. In Japanese companies, applications that automate entire workflows from data analysis to report creation are attracting attention.
Pricing Structure and Availability
The o3 model is available via API at $10 per million input tokens and $40 per million output tokens. ChatGPT Plus, Pro, and Team users can access both models without additional fees, and free users can try o4-mini with the “Think” option.
This pricing structure makes adoption in enterprise applications requiring advanced reasoning capabilities a realistic option, and is expected to advance the democratization of AI utilization.
Impact and Utilization Potential in the Japanese Market
The high-precision mathematics and science calculation capabilities of o3 and o4-mini provide significant value in Japan’s manufacturing industry, research institutions, and education sector. Particularly in complex engineering calculations, research data analysis, and educational content creation, automation at previously impossible levels has become achievable.
In the financial industry, these models’ advanced reasoning capabilities are beginning to be utilized in risk analysis and investment strategy formulation. In pharmaceutical development, applications in compound design and clinical trial data analysis are expected.
Outlook for GPT-5
GPT-5 is being developed as a true multimodal system that integrates current specialized model groups, and is planned to provide a unified AI experience that eliminates the need for traditional model switching. Its release within months is expected to bring significant changes to the competitive landscape of the AI industry.
OpenAI has also announced the release of its first open-source language model since GPT-2, and this model with reasoning capabilities is expected to further expand the base of AI development.
Future Prospects
The success of o3 and o4-mini establishes new standards for AI reasoning capabilities, enabling automation of more complex and specialized tasks. With the completion of GPT-5, the concept of AI assistants is expected to fundamentally change, making human-AI collaboration more natural and efficient.
For Japanese companies, these technological innovations represent important opportunities for strengthening competitiveness. Through early adoption and strategic utilization, establishing superiority in the global market will become possible.
Sources
- OpenAI says it’ll release o3 after all, delays GPT-5 - TechCrunch
- Introducing OpenAI o3 and o4-mini - OpenAI Official Announcement