Anthropic, OpenAI launch cheaper AI models as cost race intensifies

Anthropic and OpenAI unveiled three new AI models on September 22, intensifying the industry’s push to combine advanced capabilities with lower operating costs.

Anthropic launched Claude Opus 5.5, targeting complex coding, computer use and multi-step agentic tasks. OpenAI introduced GPT-6 Sol and GPT-6 Luna, designed to bring advanced reasoning and automation capabilities to a wider range of workloads at significantly lower prices.

Claude Opus 5.5 is aimed at tasks such as debugging large codebases, software migration, code auditing and business analysis. Anthropic said it scored 66.4 percent on Terminal-Bench 4.0, an agentic coding benchmark. Its API price is $4 per million input tokens and $20 per million output tokens, while cached input reads cost $0.20 per million tokens. Anthropic said the model is about 40 percent cheaper to operate than Opus 5 on typical workloads.

OpenAI’s GPT-6 Sol is positioned for professional work, coding, computer use and longer multi-step tasks, while GPT-6 Luna focuses on high-volume, routine workloads such as extraction, summarisation and classification. The models use improved caching and inference efficiency to reduce processing costs.

GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens. Luna is priced at just $0.10 per million input tokens and $0.50 per million output tokens, with OpenAI cutting prices by 50 percent compared with its GPT-5.6 promotional pricing.

The launches underscore a growing focus in the AI market on the cost of deploying powerful models at scale, with companies increasingly offering separate models for demanding agentic work and cheaper, high-volume applications.