GPT-5.6 Sol 性能紧逼 Fable 5 —— 但成本仅为其三分之一
gpt-5.6 sol 在智能指數上得分 59，僅比 fable 5 低一分，但每次任務成本僅為 1.04 美元，遠低於 fable 5 的 2.75 美元。這一定價策略挑戰了 anthropic 的預先市場定位。
openai released gpt-5.6 sol, a model that scores 59 on the artificial analysis intelligence index, just one point behind claude fable 5, but significantly undercuts it on price at $1.04 versus $2.75 per task. sol dominates the new coding agent index with 80 points and reduces output token usage by up to 54 percent for agentic tasks, while smaller variants terra and luna offer even lower costs. by introducing cache-read discounts and pushing a new pareto frontier for intelligence versus cost, openai is directly challenging anthropic's margins. this move heightens the price war initiated by lower-cost competitors like meta and xai, forcing the industry to prioritize economic efficiency alongside capability gains.