AI Model Makers Shift the Race Toward Lower Costs

OpenAI is releasing GPT-6 Sol and GPT-6 Luna on Tuesday, cutting costs for top business customers by 50% versus previous iterations of the same models. [1] Anthropic is releasing Opus 5.5 on Tuesday and says it costs around 40% less to run than Opus 5 while maintaining top intelligence. [1]

xAI released Grok 4.7 on Monday with a price-performance pitch, but Axios reported that new options erased much of its price advantage within 24 hours. [1] Citadel Securities found that falling per-token costs are driving additional usage while overall AI spending is increasing. [1]

Cost is becoming a more prominent competitive dimension alongside model intelligence, with the three providers emphasizing price reductions or price-performance. OpenAI's stated 50% reduction exceeds Anthropic's stated reduction of around 40%, although the figures apply to different models and descriptions of cost. Grok 4.7's quickly diminished price advantage shows how little time providers may have to capitalize on favorable economics. What the sources don't address: whether increased usage will generate enough durable revenue to offset margin pressure across model providers.