OpenAI has released two new models, GPT-6 Sol and Luna, which the company says cut API prices by up to 50 percent. The move is seen as a direct response to Anthropic's new flagship model, Claude Opus 5.5, and marks the latest escalation in the rivalry among leading AI labs. At the same time, OpenAI is pushing forward with a reform of its safety processes after a hacking incident involving a rival Anthropic model exposed fundamental vulnerabilities.
- OpenAI releases GPT-6 Sol and Luna with API pricing up to 50 percent cheaper, right after Anthropic unveiled Claude Opus 5.5.
- Three people hacked OpenAI systems using a rival Anthropic model – an incident that, according to CBS News, reveals fundamental vulnerabilities at frontier labs.
- OpenAI plans to bring in outside experts and 'watchdogs' at an earlier stage of its AI safety reviews going forward.
- Salesforce CEO Marc Benioff says Microsoft shut Salesforce out of OpenAI, so he bet on Anthropic instead – a deal that could bring in 'tens of billions.'
- According to The Verge, OpenAI is seeking advice from elite mathematicians to avoid repeating past mistakes.
What does the 50 percent price cut on GPT-6 mean for developers?
According to MarkTechPost, OpenAI has released two models, GPT-6 Sol and Luna, whose API prices are up to 50 percent lower compared to previous offerings. SiliconANGLE frames the release as an immediate response to Anthropic's Claude Opus 5.5 – OpenAI is countering its competitor with two cheaper GPT-6 variants, continuing the price pressure in the language model market.
Why is OpenAI opening its safety reviews to outside experts?
According to reports from Briefs Finance, OpenAI plans to involve outside experts at an early stage of evaluating its AI models, rather than bringing them in only late in the development process. This is a response to growing pressure on the company to supplement its internal safety checks with independent 'watchdogs' aimed at identifying potential risks early on.
The timing of the announcement coincides with a security incident: according to CBS News, three people managed to hack OpenAI systems using a rival Anthropic model. The report suggests the incident as evidence that even leading frontier labs are vulnerable to attacks via competing AI systems.
Competitive pressure is also visible beyond the technology itself: Salesforce CEO Marc Benioff said, according to TradingView, that Microsoft shut Salesforce out of a collaboration with OpenAI – in response, he bet on Anthropic instead, a wager that could bring in 'tens of billions.' The Register also describes how the pace of frontier AI development remains unabated despite calls to slow down.
What role do mathematicians play in OpenAI's new safety strategy?
According to The Verge, OpenAI wants to seek advice from elite mathematicians to avoid repeating past mistakes. The move is seen as part of broader efforts to make evaluation processes for new models more robust.
Is there resistance to the pace of AI development?
Yes. The Register reports that frontier AI labs are maintaining their development pace despite warnings. Meanwhile, according to Diya TV, a protester in San Francisco urged the city to more actively confront the risks of OpenAI's technology.
