Anthropic and OpenAI launch cheaper AI models focused on efficiency and safety

14:45

By: Azzat Manal

Anthropic and OpenAI launch cheaper AI models focused on efficiency and safety

Anthropic and OpenAI have introduced new artificial intelligence models designed to deliver advanced capabilities at lower operating costs, as competition intensifies across the AI industry. The releases come amid growing debate over the pace of frontier AI development and the need for stronger safeguards.

Anthropic unveiled Claude Opus 5.5 on September 22, describing it as a model capable of performing at the level of its higher-end Claude Fable 5.1 system on most professional tasks while costing significantly less to operate. The company said the new model requires about 40 percent less computing resources than Opus 5.

The model is aimed particularly at complex professional work, including software development, AI agents and other demanding tasks. Anthropic has also highlighted improvements in its safety and alignment systems, saying Opus 5.5 underwent external evaluation before release and includes safeguards previously used with its most capable models.

Anthropic has set the API price for Opus 5.5 at $4 per million input tokens and $20 per million output tokens. The company says the lower pricing reflects improvements in efficiency and reduced computing requirements. The model is available through Claude as well as major cloud platforms including Amazon Web Services, Google Cloud and Microsoft Azure.

OpenAI followed with the launch of GPT-6 Sol and GPT-6 Luna, two models built using techniques developed for its GPT-6 Astra system. The company is positioning the models as more affordable options for professional and large-scale workloads, with improvements in coding, factual accuracy, computer use and alignment.

GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens, while GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens. OpenAI says the prices represent a 50 percent reduction compared with the promotional pricing of its GPT-5.6 models.

The company also says the new models benefit from more efficient caching and inference, allowing them to reuse more context and respond more efficiently. According to OpenAI, Sol produces fewer factual errors than its predecessor, while both models show improvements in alignment and are less likely to misrepresent the results of coding tasks.

The releases highlight a broader shift in the AI market toward improving the cost-to-performance ratio of advanced models. Instead of relying exclusively on increasingly expensive flagship systems, companies are developing models tailored to different workloads and budgets, potentially making sophisticated AI tools more accessible to businesses and developers.

The launches also come against a backdrop of renewed discussion about AI safety. Anthropic's latest model was released after the company had called for a more measured approach to frontier AI development, while the new systems from both companies include additional safeguards intended to reduce unwanted or risky behavior.

For users and developers, the growing range of lower-cost models could expand the number of tasks that can be handled economically by AI, from software engineering and research to business operations and automated workflows. At the same time, the rapid release of new systems illustrates how quickly capabilities, pricing and safety standards are evolving across the industry.