Meta’s new ace model is extremely cost-effective! Zuckerberg returned to X after three years, and Musk quickly commented from the air

Zhidongzhi Compiler | Eggplant Editor | Cheng Qian Zhidongzhi reported on July 10 that last night, Meta Super Intelligence Laboratory released its most powerful multi-modal reasoning model Muse Spark 1.1 to date. This model is specially built for Agent tasks, focusing on improving tool calling, computer operation, programming and multi-modal understanding capabilities. It can plan tasks around user goals and call external tools to complete complex workflows. ▲Meta releases Muse Spark 1.1 (Source: X) On the same day, Meta officially launched the API public preview version of the new Meta model. Developers can call Muse Spark 1.1 through the API. This is also the first time Meta has launched a paid version to developers. The model is currently available on the Meta AI mobile app and the official website meta.ai, and users can use the model in thinking mode. Muse Spark 1.1 supports a context window of 1 million Tokens, and the input price of this model is US$1.25 per million Tokens (approximately RMB 8.49), and the output price is US$4.25 per million Tokens (approximately RMB 28.86), which is lower than many current mainstream closed-source models. ▲ Price comparison of large models (Smart East-West Tabulation) Meta founder and CEO Mark Zuckerberg also posted on X to promote the model. This is the first time Zuckerberg has posted on X in three years. Some netizens joked that they didn’t even know Zuckerberg had an X account. ▲Zuckerberg posted an article to promote Muse Spark 1.1 and netizens commented (Source: X) Musk also joined in the fun and left "Jinx" in Zuckerberg's comment area. ▲Musk’s comments (Source: X) Compared with traditional large models, which are mainly used to answer questions and generate content, Muse Spark 1.1 further strengthens the ability of AI to perform tasks. Meta said that this model can coordinate multiple agents to complete tasks together, manage 1 million token context windows, and maintain previous operations during long-term tasks.