OpenAI o3-mini Review — The Reasoning Model That Reached the Free Tier

Released January 31, 2025, o3-mini was OpenAI's first reasoning model available on ChatGPT's free tier — and the first to offer explicit reasoning effort control (low/medium/high). At high effort, it scored 87.3% on AIME 2024 and 79.7% on GPQA Diamond, matching or beating o1 at 93% lower cost. We review the architecture, benchmark breakdown by effort level, pricing, safety evaluation, and what o3-mini proved about the democratization of inference-time reasoning.