AI Price War: Claude Opus 5.5, GPT-6 Sol and Luna at Half the Cost

GPT-6 Luna costs $0.10 per million input tokens and $0.50 output, exactly half of GPT-5.6 Luna's $0.20 and $1.20, with a matching reduction in cached input pricing.

Opus 5.5 improves communication style and token efficiency while cutting pricing 20% and cache reads 60%, but its 'max' thinking level showed runaway over-thinking that can burn budget on long tasks, so teams should benchmark it on real workloads before committing.

With input tokens at $0.10 per million, long agentic conversations become dramatically cheaper, making Arabic-language AI assistants economically viable for government entities and mid-size firms, especially since cached tokens often exceed 90% of input in long sessions.

The model hit its 128,000-token output ceiling while reasoning about a trivial task, suggesting 'max' mode may be impractical for long-running work, costing time and money without guaranteeing task completion.