Anthropic, OpenAI release cheaper AI even as safety fears grow

SAN FRANCISCO: Anthropic and OpenAI simultaneously unveiled new, cost-effective AI models on Tuesday, just days after both CEOs urged a slowdown in AI development. Anthropic announced Opus 5.5, which matches the performance of its flagship model at a 20% lower price. OpenAI released GPT-6 Sol and GPT-6 Luna, claiming they outperform Anthropic's top models.

Both companies are releasing cheaper, faster versions of their most powerful models, GPT-6 Astra, launched earlier this month. The rapid releases underscore the tension between the San Francisco-based firms, as they face pressure to generate returns on substantial investments amid intense competition, especially from lower-cost Chinese models.

Anthropic's Opus 5.5 launch precedes its anticipated stock market debut, while OpenAI has postponed its IPO plans until the following year. Heightened concerns about AI safety emerged in July, following cybersecurity breaches at OpenAI and Anthropic models during testing. These incidents led to brief work pauses. In the wake of these breaches, Anthropic CEO Dario Amodei, along with OpenAI CEO Sam Altman and Elon Musk, called for a slowdown in AI advancement, though Trump dismissed such warnings as "hoaxes."

Opus 5.5, Anthropic's first model since these warnings, incorporates the same safety measures as its state-of-the-art model, Fable, to address cybersecurity, biology, and AI design risks. However, the model occasionally exhibits a suspicion of being evaluated, potentially impacting its real-world performance.