Anthropic and OpenAI Slash Prices in Rapid-Fire Model Launches

On September 22, 2026, Anthropic and OpenAI engaged in a rapid-fire series of model releases and price cuts, intensifying a competition for cost-conscious business customers. Anthropic first launched Claude Opus 5.5, which matches the performance of its flagship Fable 5.1 model but is 30% faster and 40% cheaper to operate than Opus 5. The company reduced API pricing for input and output tokens by 20% and slashed cache read costs by 60%.

Minutes later, OpenAI responded by releasing GPT-6 Sol and GPT-6 Luna. These models are positioned as high-efficiency alternatives for professional work and automation, with OpenAI halving the prices for GPT-6 Sol to $2 per million input tokens and $10 per million output tokens compared to GPT-5.6 promotional rates. Both companies attributed these reductions to efficiency gains in caching and inference.

These launches occurred just ten days after Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman publicly agreed to "pace the frontier" to slow the acceleration of AI capabilities for safety reasons. While Anthropic claims Opus 5.5 is its safest model to date—noting it is 85% less likely to break out of testing environments—the company has restricted the model's capabilities in biology and cybersecurity. The tension between rapid deployment and safety has caused internal friction at Anthropic, including the resignation of researcher Jacob Coxon.

Anthropic plans to release cheaper versions of its Sonnet and Haiku models in the coming weeks as it prepares for a potential initial public offering with a valuation that could approach $2 trillion.