Anthropic, OpenAI Cut AI Model Costs as Opus 5.5 Prices Fall 20%
Updated
Updated · Ars Technica · Sep 22
Anthropic, OpenAI Cut AI Model Costs as Opus 5.5 Prices Fall 20%
3 articles · Updated · Ars Technica · Sep 22
Summary
Anthropic unveiled Opus 5.5 while OpenAI rolled out GPT-6 Sol and Luna, extending a same-day push by both companies toward cheaper, faster enterprise AI rather than major capability leaps.
Enterprise demand is driving the shift as customers increasingly use model routers and open-weight alternatives to limit reliance on expensive frontier models.
Opus 5.5 is Anthropic’s clearest price move: input and output tokens cost $4 and $20 per million, 20% below Opus 5, while cache reads drop 60% to $0.20 per million.
Anthropic also said Opus 5.5 generates output more than 30% faster and modestly outperforms OpenAI’s GPT-6 Astra on some coding and knowledge-work benchmarks.
The releases sharpen a pricing battle already underway after OpenAI earlier launched Sol and Luna at roughly half the cost of its 5.6 series and added GPT-6 caching discounts of up to 90%.