OpenAI has introduced its new artificial intelligence model, o3, as the successor to its previous model, o1. Alongside the primary version, the company also unveiled a compact modification, o3-mini, designed for specialized tasks. This development paves the way for new possibilities in simulating cognitive processes, bringing technology closer to the level of artificial general intelligence (AGI). Interestingly, OpenAI skipped the o2 name to avoid associations with the UK telecom operator O2.
According to the company, the o3 model demonstrates impressive results and, under certain conditions, approaches AGI. However, OpenAI emphasizes that this is not yet the ultimate breakthrough in the field. In the ARC-AGI test, which measures an AI's ability to learn beyond initial data, o3 achieved an 87.5% score in high-performance computing mode, nearly tripling the results of the o1 model.
One of the key innovations in the o3 model is the ability to adjust performance levels, allowing users to choose between low, medium, or high computational modes depending on task complexity. The high-performance mode delivers the best results but requires more processing time—ranging from a few seconds to several minutes.
Building on the experience of the previous model, OpenAI acknowledges potential risks associated with the use of o3. To mitigate these risks, initial access to o3-mini will be granted to security researchers. The full-scale release of o3 is planned for 2025.
Would you like your employees to receive real-time data on new Facebook leads, and automatically send a welcome email or SMS to users who have responded to your social media ad? All this and more can be implemented using the SaveMyLeads system. Connect the necessary services to your Facebook advertising account and automate data transfer and routine work. Let your employees focus on what really matters, rather than wasting time manually transferring data or sending out template emails.