The Official Launch

OpenAI officially launched o3 for general availability on January 10, 2026, following the November 2025 benchmark preview and a limited safety testing period. The launch introduced two variants: o3-mini (optimised for speed and cost, with strong STEM reasoning) and o3 (the full model with the highest capability). Both were available via the OpenAI API and in ChatGPT, with access tiers based on subscription level.

o3-mini: The Developer Model

o3-mini at $1.10/million input tokens and $4.40/million output tokens was positioned as the practical reasoning model for most API use cases. It supported three reasoning effort settings — low, medium, and high — allowing developers to trade response time for reasoning depth. At low effort, o3-mini responded in 15-30 seconds; at high effort, it could reason for several minutes for the most complex problems.

o3 Full: The Pro Tier

o3 full was available to ChatGPT Pro subscribers and via API at $15/million input, $60/million output. The full model demonstrated capabilities that remained beyond o3-mini: genuine novel mathematical problem solving, complex multi-step scientific reasoning, and software architecture decisions that required understanding large, interrelated system designs. On ARC-AGI, o3 full scored 87.5% at high compute — still the highest score of any AI system on this benchmark.

Safety and Limitations

OpenAI conducted extensive red-teaming before the o3 launch, noting that the model's improved reasoning capabilities also made it better at reasoning about potentially harmful scenarios — requiring additional safety fine-tuning. The company implemented usage monitoring for high-risk task categories and maintained human review processes for certain sensitive use case categories.

What This Means for Indian Businesses

OpenAI o3's official launch at the start of 2026 makes frontier reasoning AI available to Indian developers via API for the first time. o3-mini at $1.10/million input tokens is affordable for production use cases requiring deep reasoning: legal analysis, financial modelling, scientific literature review, and complex debugging. Indian AI startups building in these domains should evaluate o3-mini as a core model for their most demanding tasks.