ทันเอไอ

Meta Launches Muse Spark 1.1, A New AI Coding Model, Challenging OpenAI and Anthropic

Meta escalates competition in the developer AI market with the launch of Muse Spark 1.1, a language model focused on coding and automation capabilities. It is now available to US developers for testing via API.

Just three months after officially launching its own AI model, Meta has released a major update named Muse Spark 1.1. The company describes it as a 'step-change' from the first version, directly targeting the AI coding tools market to compete with rivals like OpenAI and Anthropic. Alexandr Wang, Meta's Head of AI, told CNBC that Muse Spark 1.1 is currently the company's 'strongest model for agentic and coding tasks.' 'Agentic' tasks refer to the AI's ability to perform complex, multi-step operations autonomously. Wang also claimed the new model outperforms competitor models in certain tasks involving interaction with third-party coding services and software. However, these claims come from Meta's own executives, and independent comparative test results have yet to be released. The most significant move is the broadened access. While the initial model in April was available only to invited partners via a private API preview, Muse Spark 1.1 is now accessible to general developers in the US who can register through a public preview via the developer portal. This clearly signals Meta's readiness to attract more developers to its platform. Furthermore, Wang revealed that the team is training an even more powerful next-generation AI model codenamed 'Watermelon,' though no clear launch timeline has been specified. Previously, Business Insider reported that Meta's new model might be comparable to GPT-5.5 in certain benchmarks, but the report did not disclose which specific tests were involved.