OpenAI and Anthropic unveil new generation of artificial intelligence

Anthropic and OpenAI have almost simultaneously unveiled new AI models: Claude Opus 5.5 and GPT-6 Sol/Luna. The companies are reducing API costs and improving efficiency, shifting focus from a power race to a battle over price and performance. Independent tests show Anthropic leading in programming, while OpenAI leads in business processes.

On September 22, Anthropic and OpenAI almost simultaneously unveiled the next generation of their artificial intelligence models. Anthropic launched Claude Opus 5.5, while OpenAI launched GPT-6 Sol and GPT-6 Luna. Both companies significantly reduced API costs: Anthropic's cost is down to $4 per million input tokens and $20 per million output tokens, while OpenAI reports a reduction of about 50% compared to previous generation prices. Anthropic also solved the "Claudish" problem—the model's tendency toward formulaic phrasing. According to independent tests, Claude Opus 5.5 scored 58 points in the Intelligence Index ranking, taking first place, and showed 67.7% accuracy in Humanity's Last Exam versus 57.2% for GPT-6 Astra. In programming, Opus 5.5 scored 66.4% in Terminal-Bench 4.0 versus 57.9% for GPT-6 Astra. However, in business processes (AutomationBench), GPT-6 Astra scored 41.4% versus 40% for Opus 5.5, and in scientific research (Terminal-Bench-Science), it scored 64.6% versus 58.7%. The article notes that price reductions are driven by competition, especially from Chinese models, and regulatory pressure, including the European AI Act. For Israeli companies, this opens opportunities but requires a review of security protocols.