Doha:Anthropic and OpenAI have launched new, more affordable artificial intelligence models almost simultaneously, following recent calls from both companies to slow down AI development.
According to Qatar News Agency, Anthropic unveiled its Opus 5.5 model, which offers near-equivalent performance to its flagship model at a significantly reduced price. Opus 5.5 is priced 20% lower than Anthropic's advanced model, Fable, while delivering comparable results on most tasks.
OpenAI soon introduced GPT-6 Sol and GPT-6 Luna, claiming these models outperform Anthropic's top models. These new models are more economical and faster variants of GPT-6 Astra, OpenAI's leading model released earlier in September.
The launches highlight the tension between the two companies as they express concerns over AI's potential dangers while striving to generate returns on substantial investments. They face intense competition from lower-cost models, particularly those from China.
Anthropic's model release comes in anticipation of its expected stock market listing. In contrast, OpenAI has delayed its initial public offering plans until 2027. Despite the rush to market, Anthropic assures that Opus 5.5 employs the same security measures as its Fable model in cybersecurity, biology, and AI design.
Anthropic also reported that Opus 5.5, based on internal tests, is their best-performing model concerning alignment, meaning it adheres closely to designer intentions. However, the model sometimes appears to be aware of being evaluated, which complicates predictions of its behavior in real-world applications.
