OpenAI and Anthropic Race to Cut Costs as Enterprises Shift to Cheaper Models

OpenAI and Anthropic have both unveiled new model releases designed to lower operational expenses rather than push performance boundaries. Anthropic's Opus 5.5 and OpenAI's GPT-6 Sol and Luna represent a strategic shift toward cost efficiency—a response to enterprise customers increasingly exploring alternatives and deploying router systems that favor cheaper models over premium offerings.

The Cost-Efficiency Imperative#

The competition for enterprise dollars is intensifying as organizations reassess their AI spending. Rather than chasing marginal gains in capability, both companies now face pressure from cheaper open-weight models that perform adequately for many real-world use cases. By substantially reducing pricing, Anthropic and OpenAI are attempting to retain customers who might otherwise shift to cost-effective alternatives.

Opus 5.5: Modest Gains, Major Savings#

Anthropic's updated flagship model targets users performing coding and complex knowledge work. While the performance improvements over its predecessor are incremental—showing only modest advantages over OpenAI's recently released GPT-6 Astra in certain benchmarks—the pricing changes are substantial.

According to Anthropic's announcement, input and output tokens now cost $4 and $20 per million tokens respectively, representing a 20% reduction. More significantly, cached token reads—which dominate agentic and coding workloads—dropped 60% to $0.20 per million tokens. The model also delivers over 30% faster output generation.

The Broader Pattern#

These releases signal that the frontier of AI competition has shifted from raw capability to practical economics. As enterprises build more sophisticated workflows and routing logic to optimize costs, the premium models must justify their price tags through either clear performance superiority or substantially lower expenses. Both Anthropic and OpenAI are betting the latter will prove decisive.