Meta 发布 Muse Spark 1.3 多基准测试领先且成本更低
Meta 新模型 Muse Spark 1.3 以更强性能和近零成本冲击市场，可能改变 AI 价格竞争格局。
Meta 发布新一代模型 Muse Spark 1.3，通过减少无效交互和精简输出，大幅降低 token 消耗。它在 DeepSWE 等长程编程基准上超越 GPT-5.6 Sol 和 Claude Opus 5，成本却极低——一次 2 小时的复杂任务不到 1 美元。模型定价未涨，输入每百万 token 1.25 美元，输出 4.25 美元。
正文摘录
 Report from AI Era (新智元)  The competition is brutal. September is only three days old, and already five frontier models have been released! Anthropic's Fable 5.1 and Mythos 5.1, Google's Gemini 3.8 Flash and Cyber, and Meta's Muse Spark 1.3... RSI (the AI race) has definitely arrived. Just last night, Google's Gemini 3.8 Flash became the lightweight champion. Then Meta came to crash the party. Zuckerberg made a bold announcement on X: Muse Spark 1.3 launches today, delivering frontier performance at a price so low it's practically unmetered. This is our biggest leap forward in coding and agentic work yet! ![](https://aiera.com.cn/wp-cont…