OpenAI and Anthropic Cut AI Model Costs by Up to 80%
OpenAI and Anthropic have lowered costs for some AI models and API workloads, as competition intensifies for developers and businesses using generative AI.
The largest headline reduction came from OpenAI: in July, it cut the API price of GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%. Anthropic’s Claude Fable 5.1 brought a separate reduction in cached-input pricing, with the company estimating that this could lower costs by around 25% for typical workloads and up to approximately 45% for highly agentic tasks.
The figures describe different pricing changes and should not be read as an 80% reduction across all models or products.
OpenAI’s cuts apply to two API models
OpenAI said the price cuts followed improvements to how its models are served. From July 30, the API rate for GPT-5.6 Luna fell to $0.20 per million input tokens and $1.20 per million output tokens. GPT-5.6 Terra’s rates fell to $2 per million input tokens and $12 per million output tokens.
The 80% reduction applied to Luna, while Terra’s reduction was 20%. OpenAI said prices for GPT-5.6 Sol were unchanged in that update. It also said ChatGPT and Codex subscription prices and quota budgets remained unchanged, although Luna and Terra usage counted for fewer credits.
Anthropic reduced cached-input costs
Anthropic’s Fable 5.1 update reduced the price of cache reads by 75%, to $0.25 per million tokens. Cache reads occur when a model reuses input it has already processed.
Anthropic said Fable 5.1 costs about 25% less than Fable 5 for typical usage-based workloads, and up to around 45% less for highly agentic work. The actual savings depend on how a customer’s workload uses cached input; the model’s base input and output rates were not cut by 75%.
What the reductions mean for AI users
Lower API rates can reduce costs for developers and companies that pay based on model usage. The effect on an individual bill depends on which model is used, the number of input and output tokens, and how much context can be reused through caching.
The announced changes do not mean that every AI plan became cheaper. OpenAI said its subscription prices were unchanged, while Anthropic’s cited savings describe usage-based model costs. Buyers comparing services should check current rates and estimate their own workload rather than treating the largest percentage as a universal discount.