OpenAI to Launch o3 Mini: A Faster, Cheaper AI Reasoning Model in September 2026

OpenAI announced it will release o3 Mini, a compact AI reasoning model, on September 12, 2026, with on-premise licensing to follow in Q4 2026. Built on a 2.3-billion-parameter architecture, the model runs up to three times faster and consumes 70% less energy per token compared to its predecessor, o3 Standard. Despite its smaller size, OpenAI claims o3 Mini performs within 2% accuracy of its larger counterparts across reasoning-heavy tasks such as code analysis and complex problem solving. Priced at $0.001 per 1,000 tokens — roughly 30% cheaper than the current o3 model — the release targets startups, SMBs, and independent developers who lack access to large GPU infrastructure. OpenAI has partnered with Qualcomm to test edge deployment on the Snapdragon X Elite platform, and the company hinted at future expansions including an ultra-compact o3 Nano model and a multimodal variant.