A rapid series of price cuts has transformed the large language model market in under six months. OpenAI and Anthropic have engaged in aggressive pricing, reducing costs by up to 80% to maintain competitiveness. In June 2026, OpenAI launched GPT-5.6 Sol and Luna, with Luna positioned as a cheaper alternative.
Anthropic followed with Claude Sonnet 5 and Opus 5, highlighting improved performance. By July, OpenAI slashed Luna's price by 80% after benchmark comparisons. August saw Anthropic making its Sonnet 5 price reduction permanent.
September brought further reductions: OpenAI cut GPT-5.6 Sol by 20%, while Anthropic reduced cache read costs for Fable 5.1 by 75%. The most significant drop occurred on September 22, when OpenAI launched GPT-6 Sol and Luna at half the price of the discounted GPT-5.6. On the same day, Anthropic released Opus 5.5, noting a 20% input/output cost cut and larger cache read reductions.
By September 29, OpenAI halved GPT-6 Sol's cache price, making the model family 75% cheaper overall. These moves signal a price war driving down costs for users, with cache read costs emerging as a critical factor in total LLM expenditure for agentic coding workflows.
Source: Hacker News · Summarized by HeadlinesBriefing