OpenAI has unveiled its latest GPT-5.5 model for subscribers who pay for access. The developers assert that this version represents the pinnacle of AI capabilities, particularly excelling in programming, scientific inquiries, workplace tasks, and agent-like functions. These claims are supported by various benchmarks and independent assessments. The GPT-5.5 achieved a score of 60 on the AI Intelligence Index from Artificial Analysis, surpassing its main competitors, Claude Opus 4.7 and Gemini 3.1 Pro. However, the new model comes with a price tag that is twice that of its predecessor. The API pricing is set at $5 for input tokens and $30 for output tokens per million, making it the most expensive option among leading AI models, even exceeding the cost of Opus 4.7. On a positive note, GPT-5.5 demonstrates impressive token efficiency, which helps mitigate the higher expense. Nevertheless, experts from Artificial Analysis have pointed out a concerningly high rate of hallucinations during their evaluations—despite GPT-5.5 having the most robust knowledge base of any existing AI. This model fabricates answers instead of acknowledging a lack of knowledge approximately 1.5 to 2 times more frequently than its competitors, Opus 4.7 and Gemini 3.1 Pro. The release schedule for cutting-edge AI models is still ongoing. Elon Musk's xAI is currently testing Grok 4.3, and the Chinese lab DeepSeek is anticipated to introduce the highly awaited V4 shortly. Consequently, OpenAI's claim of having "the smartest model" may soon face significant challenges from emerging competitors.
Informational material. 18+.
