OpenAI and Anthropic intensify price competition in the AI market

OpenAI and Anthropic almost simultaneously introduced new artificial intelligence models with significantly lower usage costs. The two largest competitors in the generative AI market are increasingly competing not only on model performance but also on the cost of their use by businesses and developers.

On September 22, Anthropic introduced Claude Opus 5.5. About an hour and a half later, OpenAI announced the release of GPT-6 Sol and GPT-6 Luna, TechCrunch notes.

Anthropic positions Opus 5.5 as its strongest model in the Opus lineup. Input tokens cost $4 per million, output tokens $20. For the previous Opus 5, they cost $5 and $25, respectively.

However, the company estimates the total cost reduction for a typical task at about 40%. This is due not only to cheaper tokens but also to the fact that the new model uses fewer tokens to perform the same work. Cache read costs have also decreased from $0.50 to $0.20 per million tokens.

Almost simultaneously, OpenAI released GPT-6 Sol and GPT-6 Luna. The company claims that thanks to improved caching and computation efficiency, API prices for the new models are 50% lower compared to the promotional rates of GPT-5.6 Sol and Luna.

At standard pricing, GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens. The significantly more compact GPT-6 Luna is aimed at mass and relatively simple tasks – its cost is $0.10 per million input and $0.50 per million output tokens.

OpenAI targets Sol primarily at complex programming and AI agent work, while Luna is designed for large volumes of more clearly defined tasks – for example, document summarization, information extraction, and quick responses.

Companies are simultaneously trying to reduce the cost of using their most performant models. Anthropic notes that Opus 5.5 performs at the level of Claude Fable 5.1 in most tasks, while being cheaper to run. OpenAI, in turn, compares GPT-6 Sol with Anthropic's leading models and claims improvements in programming and factual accuracy.

Rate reductions are especially important for companies that run large volumes of requests through AI or use agentic systems capable of autonomously performing long sequences of operations. In such scenarios, even a small difference in the cost per million tokens can significantly affect overall expenses.

GPT-6 Sol and Luna are now available through the API, as well as in OpenAI products. Claude Opus 5.5 is available through the Claude Platform, as well as cloud platforms Amazon Web Services, Google Cloud, and Microsoft.

Based on materials from: Anthropic, OpenAI, TechCrunch