OpenAI's latest reasoning model, o3, outperforms its predecessor o1 in key cognitive areas, enhancing human-like intelligence for coding, math, and concept reasoning.
O3 incorporates a safety feature known as deliberative alignment, which employs a 'chain of thought' approach to mitigate risks associated with jailbreaking and user manipulation.
The launch of o3 coincided with OpenAI's '12 Days of OpenAI' campaign, showcasing a range of innovations including a high-tier subscription service for ChatGPT users.
OpenAI has also announced a Reinforcement Fine-Tuning Research Program aimed at allowing developers to create custom models tailored for specific applications, set for alpha testing in 2025.
Collection
[
|
...
]