Anthropic has released a new, cost-effective AI model. How did the creator of ChatGPT respond?

Anthropic, the world’s most valuable AI startup, released a new artificial intelligence model, Claude Opus 5.5, on September 22. According to the developer, it is capable of performing most tasks at the level of the company’s most powerful model, Fable 5.1, but costs users significantly less than its predecessor, Opus 5. On the same day, Anthropic’s main competitor—OpenAI— launched new versions of its cutting-edge models: GPT-6 Sol and GPT-6 Luna.

Anthropic’s Opus 5.5 costs $4 per 1 million input tokens and $20 per 1 million output tokens. With standard settings, this is 40% cheaper than the previous generation of this model, Opus 5, the company claims. By comparison, its most advanced model, Fable 5.1, costs $10 and $50 per 1 million tokens, respectively, notes the Financial Times. At the same time, according to Anthropic, the new model is capable of performing most tasks at the same level. Opus 5.5 marks the launch of a new family of models that will be more cost-effective than previous generations, the company said.

OpenAI claims that it tested GPT-6 Sol and GPT-6 Luna using methods similar to those used to test its flagship model, Astra, which it calls the smartest in the world. However, Sol and Luna are faster and, at the same time, less expensive, the company noted. For GPT-6 Sol, an input token will cost $2, and an output token will cost $10. This is 50% cheaper than the previous generation of this model.

Opus 5.5 was Anthropic’s first release after the company’s CEO, Dario Amodei, called on AI labs to slow down the pace of development of cutting-edge models. In his view, this will help reduce the risk of security breaches following warnings from researchers that AI could harm humanity. Anthropic states that the Opus 5.5 model underwent extensive testing to ensure compliance with safety principles, including an independent assessment by third-party organizations prior to release. The new AI also features safeguards in high-risk areas such as cybersecurity and biology.

Investors believed that the launch of the new model would serve as a catalyst for the publication of Anthropic’s long-awaited IPO prospectus, according to the Financial Times. They are closely watching to see whether the startup can maintain its high growth rate in the coming years amid intensifying competition and growing price sensitivity among customers, the publication notes. The company, whose customer base consists primarily of corporate clients, is seeking to expand it ahead of the offering.

Anthropic is responding to intensifying competition by lowering its prices. OpenAI has also been cutting the cost of its models in recent months in an effort to accelerate growth following a slow start to the year. Open models—primarily from China—are putting additional pressure on both companies: users can configure and run them on their own hardware. According to the cloud company Vercel, these models are taking market share away from American developers.

According to Bloomberg Intelligence, Anthropic’s annual recurring revenue reached approximately $47 billion by June, more than quintupling since December. In May, the company closed a funding round with a valuation of $965 billion, surpassing OpenAI’s $852 billion valuation. Anthropic filed a confidential application for an IPO in the U.S. on June 1, and a week later, OpenAI took a similar step.

This article was AI-translated and verified by a human editor