Anthropic, OpenAI Launch Cheaper AI Models Despite Mounting Safety Concerns

Anthropic and OpenAI have rolled out new, less expensive artificial intelligence models within hours of one another, just days after executives from both companies called for a slowdown in AI development.

Anthropic said its new Opus 5.5 model comes close to matching the performance of its flagship Fable model on most tasks, while costing 20 percent less. OpenAI followed shortly after with two new models, GPT-6 Sol and GPT-6 Luna, which it said handle tasks substantially better than competing offerings.

The new OpenAI models are cheaper, faster variants of GPT-6 Astra, the company's most powerful model, which launched earlier in September.

The near-simultaneous launches highlight a contradiction both companies face: even as their executives warn publicly about the risks posed by the technology they are building, both firms remain under pressure to generate returns on massive investments amid stiff competition from lower-cost rivals, particularly those emerging from China.

Anthropic's launch comes just weeks ahead of its expected stock market listing, while OpenAI has pushed back its own initial public offering plans to 2027.

Anthropic said the new model applies the same safeguards as its state-of-the-art Fable model in the areas of cybersecurity, biology, and AI model design.

The company also said that, based on its own internal testing, Opus 5.5 is the best-performing model it has evaluated to date on alignment — meaning how closely the model's behavior adheres to what its designers intend.

However, Anthropic acknowledged that the model also showed signs of frequently suspecting it was being evaluated during testing, a tendency the company said makes it harder to predict how it will behave in real-world use. 

Source : QNA