OpenAI researchers say they've discovered hidden features inside AI models that correspond to misaligned "personas," ...

Researchers at the company looked into how malicious fine-tuning makes a model go rogue, and how to turn it back.

According to a recent study by OpenAI, AI models can develop personas of their own. AI models with their own personas The ...

Every few months, an AI lab in China that most people in the U.S. have never heard of releases an AI model that upends ...

The latest version of OpenAI’s most intelligent AI model, o3-pro, outperforms previous models on benchmarks for math, science ...

OpenAI claims that o3-pro exhibited significantly improved performance over o3 in evaluations regarding science, education, ...

OpenAI has phased out its partnership with Scale AI as part of a long-term strategy to find data suppliers more aligned with ...

OpenAI has been teasing news of a new open-weight model for a while now, originally planning for an early summer release.

Despite the strong reaction, OpenAI had in fact already announced the plan to deprecate GPT-4.5 Preview back in April 2025.

It could benefit startups, research teams, and individual developers who previously found higher-tier model access ...

B-Preview, an open source AI coding model based on Deepseek-R1-Distilled-Qwen-14B. The model achieves a 60.6% pass rate on ...

Learn more about why OpenAI is delaying its open-weight AI model, including an X post from Sam Altman about "something ...