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

The July 30 cut moved GPT-5.6 Luna from $1 / $6 / $0.10 down to $0.20 / $1.20 / $0.02. OpenAI tied the move to efficiency gains it found after GPT-5.6 helped make itself cheaper to run. Anthropic's Sonnet 5 launched with launch pricing of $2 / $10 / $0.20 which was initially supposed to be temporary, and that rate was later kept in place. The author argues that cheap open-weight models and many competing inference providers set the ceiling on closed-model pricing. DeepSeek's original DeepSeek V4 Flash 0420 was released in April 2026 at a price of $0.14 / $0.28 / $0.0028. The post also says that without the Luna cut, OpenAI would have lost to DeepSeek V4 Pro on cost per task. In September, GPT-6 Sol was priced at $2 / $10 / $0.20, and Fable 5.1 cut the cache read cost by 75%. For a team building an AI company, cache reads matter most in agent workloads. The post estimates that a 75% reduction in the cache read cost correlates to a 70-75% reduction in the cost of using the model. Flagship pricing moved less than the cheaper tiers, since GPT-5.6 Sol stayed at $5/$30, so model routing and margin assumptions need rechecking.