A look at OpenAI's and Anthropic's aggressive price cuts between June and October, driven by low switching costs and heavy competition from open-weight models

The cost to use leading large language models has fallen by up to 75% in just four months.

Between June and October 2026, OpenAI and Anthropic implemented significant price reductions on their large language models (LLMs). OpenAI's GPT-5.6 Luna saw an 80% price cut, dropping to $0.20 per million input tokens, $1.20 per million output tokens, and $0.02 per million cache reads. Anthropic adjusted its Claude Sonnet 5 pricing and later cut the cache read cost for Fable 5.1 by 75%. By late September, OpenAI announced GPT-6 Sol and Luna at halved prices compared to their GPT-5.6 counterparts, and Anthropic released Claude Opus 5.5 with a 40-60% total cost reduction. These aggressive cuts are attributed to intense competition from open-weight models and low switching costs for users employing LLM-based agents.

The rapid price reductions by major LLM providers like OpenAI and Anthropic indicate intense market pressure, primarily from the proliferation of competitive open-weight models. Companies are forced to lower prices to retain market share, as the ability for users to switch between models with minimal effort has increased. This price war suggests a shift towards LLMs becoming more commoditized, where differentiation must come from superior performance or unique features rather than price alone. The significant drops, especially in cache read costs which are crucial for agentic applications, highlight the need for providers to innovate beyond raw token pricing.

This trend of aggressive price cuts by closed-model providers is directly challenging the business models of companies relying on premium pricing. Open-weight models, with their lower overhead and increasing accessibility through inference providers, are setting new cost benchmarks. As a result, OpenAI and Anthropic are likely to continue investing heavily in R&D to maintain a performance edge, while also exploring alternative revenue streams such as enterprise-specific features or specialized agentic services to build stronger customer lock-in. The market is moving towards a state where raw LLM compute will become increasingly inexpensive, shifting value to application-layer innovation and integration.