OpenAI and Anthropic are aggressively lowering prices for their mid-tier AI models as corporate clients struggle with rapidly depleting technology budgets. This shift occurs as both companies prepare for potential initial public offerings, forcing them to balance revenue growth with competitive pricing strategies.
In late July 2026, OpenAI reduced the price of its GPT-5.6 Luna model by 80%, setting costs at $0.20 per million input tokens and $1.20 per million output tokens. Anthropic followed in August 2026 by setting Sonnet 5 pricing at $2 per million input and $10 per million output tokens.
Anthropic also lowered the cost of its Opus 5.5 model to $4 per million input and $20 per million output tokens, achieving nearly 40% savings. In early September 2026, the company further reduced cache-read pricing for its Fable 5.1 model by 75% to $0.25 per million tokens within a 72-hour window.
Market data indicates that blended token prices have declined by 41% to 64% since mid-2026. While token volume has increased 25-fold since 2023, enterprises are now implementing spending caps and shifting workloads to mid-tier models, which currently provide approximately 90% of flagship performance.
