What once seemed like a frantic race between Anthropic and OpenAI to one-up each other with new AI models is now running under the caution flag, thanks to the US government's decision to insert itself into the release pipeline citing concerns about AI safety.
Three new models in OpenAI's GPT-5.6 series arrived Friday as a "limited preview" available to "a small group of trusted partners whose participation has been shared with the government," the company said in a blog post. Bloomberg reported that 20 partners have been granted access to the models through AWS's Bedrock service.
The new flagship model is called Sol, and OpenAI said it outperforms Anthropic's Claude Mythos 5 on coding tasks while emphasizing that it "strengthened protections for higher-risk activity, sensitive cyber requests, and repeated misuse, and spent multiple weeks finding weaknesses, pressure-testing our system, and hardening it against real-world attacks" before releasing the preview. The other two models are called Terra, which OpenAI said compares to the GPT 5.5 series models in performance but for half the cost, and Luna, which is the budget option.
