OpenAI o3 receives 80% price cut, premium o3-pro model launched simultaneously

On June 10, 2025, OpenAI simultaneously announced a significant price reduction for its reasoning model o3 and the release of the higher-performance o3-pro¹. This move could represent a major turning point for developers amid intensifying price competition in AI models.

CEO Sam Altman announced on X (formerly Twitter) “we cut o3 pricing by 80%!!” with new pricing set at $2 per million input tokens and $8 per million output tokens¹. This represents exactly an 80% reduction from the previous pricing of $10 input and $40 output.

Details and significance of o3 price reduction

Major pricing revision details

Through infrastructure optimization of the o3 model, OpenAI can now offer the same model at significantly lower costs. The new pricing structure is as follows:

Standard mode: $2 per million input tokens, $8 per million output tokens
Flex mode (synchronous processing): $5 per million input tokens, $20 per million output tokens¹
With caching feature: Additional $0.50 discount per million tokens

This pricing revision makes o3 equivalent in token pricing to GPT-4.1 and cheaper than GPT-4o. OpenAI particularly recommends trying o3 for coding tasks, agent tool calls, function calling, and instruction-following tasks.

Enhanced competitiveness

This significant price reduction positions o3 to directly compete with Google’s Gemini 2.5 Pro and Anthropic’s Claude models on pricing². Sam Altman emphasizes that this new pricing is intended to facilitate broader experimentation.

For developers, advanced reasoning capabilities of o3 can now be utilized at practical costs. Particularly expected for applications requiring complex logical thinking or mathematical reasoning.

Introduction of premium o3-pro

o3-pro features and performance

o3-pro was released on the same day as the premium version of o3. By using more computational resources to “think deeper,” it provides reliable answers to difficult problems³.

According to OpenAI’s announcement, in expert evaluations, o3-pro showed superior results to o3 across all test categories, with particularly notable performance improvements in science, education, programming, business, and writing assistance³.

Benchmark performance

o3-pro’s benchmark results are very impressive:

AIME 2024 (mathematical skills assessment): Score surpassing Google’s highest-performance model Gemini 2.5 Pro
GPQA Diamond (doctoral-level scientific knowledge test): Results exceeding Anthropic’s latest model Claude 4 Opus³

These results demonstrate that o3-pro has the highest performance among currently available AI models.

Pricing and availability

o3-pro is priced at $20 per million input tokens and $80 per million output tokens³. While this is 10 times the price of standard o3, it represents an 87% price reduction compared to the previous o1-pro.

For availability, it will first be selectable in the model picker for Pro and Team users, replacing o1-pro. Enterprise and Edu users will have access starting the following week³.

Technical details and usage methods

API usage

o3-pro is available in the Responses API as ‘o3-pro-2025-06-10’ and supports the following features:

Image input
Function calling
Structured Outputs

However, since o3-pro is designed to handle difficult problems, some requests may take several minutes to complete. Using the new background mode of the Responses API is recommended to avoid timeouts.

Recommended use cases

OpenAI recommends using o3 and o3-pro for the following types of tasks:

Coding tasks (o3 particularly advantageous price-wise)
Agent tool calls
Function calling and instruction following
Complex scientific reasoning (o3-pro)
Advanced mathematical problem solving (o3-pro)

Impact on Japanese development landscape

Cost benefits

The 80% price reduction of o3 enables Japanese companies and developers to utilize advanced reasoning models at practical costs. Particularly for startups and SMEs, access to cutting-edge AI technology that was previously cost-prohibitive has significantly improved.

For example, development of tools leveraging o3’s advanced reasoning capabilities in areas such as code review, debugging assistance, and architecture design is expected to accelerate.

Changing competitive environment

With intensifying price competition against Google’s Gemini 2.5 Pro and Anthropic’s Claude, Japanese companies can now compare and evaluate multiple AI providers, selecting optimal models based on specific use cases. This enables cost-effective AI utilization while avoiding vendor lock-in.

Premium model utilization scenarios

With the introduction of o3-pro, Japanese research institutions and R&D departments of large corporations are expected to utilize it for more advanced problem-solving. Particularly in fields such as materials science, pharmaceuticals, and financial engineering, o3-pro’s doctoral-level scientific knowledge and reasoning capabilities may find application.

This announcement will likely serve as an important milestone demonstrating the democratization of AI technology and intensification of competition. Further price competition and technological innovation are expected to accelerate, requiring Japanese companies to actively respond to these trends.