OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?

LLM API calls are now significantly cheaper, reducing the cost of building and running AI-powered applications.

OpenAI and Anthropic have drastically reduced the prices of their large language models (LLMs) over a few months, with cuts reaching up to 80%. The price reductions began with OpenAI's GPT-5.6 Luna, which saw an 80% price drop, followed by Anthropic making its Claude Sonnet 5 launch pricing permanent. OpenAI then reduced prices for GPT-5.6 Sol, and Anthropic launched Claude Opus 5.5 with significant cost cuts. These moves appear to be a direct response to competitive pressures, particularly from open-weight models like DeepSeek Flash V4.1, which offer comparable performance at lower costs. For instance, DeepSeek Flash V4.1's cache read costs are substantially lower, making its effective cost even less than advertised. The trend indicates a race to the bottom in pricing, driven by increased competition and the commoditization of LLMs, forcing major players to aggressively cut costs to remain competitive, especially as new models like Moonshot AI's Kimi K3.1 Pro emerge with competitive performance at lower prices.

The aggressive price cuts from OpenAI and Anthropic signal a significant shift towards commoditization in the LLM market. Increased competition from open-weight models and numerous inference providers has eroded the moats previously held by proprietary models. This dynamic forces major players to engage in price wars, akin to the early days of ride-sharing services, to retain market share. As a result, the effective cost of using advanced AI capabilities is plummeting, making AI more accessible for a wider range of applications and developers.

The sustained price reductions, especially in cache read costs, suggest that the underlying inference technology is maturing and becoming more efficient. However, the intense competition also pushes companies like OpenAI and Anthropic to innovate rapidly and build new forms of differentiation, such as enhanced enterprise features or specialized agentic assistants. Developers should anticipate continued price volatility and explore models that balance cost, performance, and specific feature sets as the market continues to evolve.