Meta Platforms (META.US) releases the new generation language model Muse Spark 1.1, adding AI intelligence to the code generation market.

Just two days after launching the AI image generation model Muse Image, Meta Platforms (META.US) once again upgraded its artificial intelligence (AI) product matrix. On Thursday, Meta released the new generation large language model Muse Spark 1.1, and opened its API for the first time in a public preview form, further enhancing its AI intelligence and code generation market, competing head-on with rivals such as OpenAI and Anthropic. After the news was announced, Meta's decline narrowed to 0.5%, after falling by about 4% earlier. Alexandr Wang, head of Meta's Super Intelligence Lab, said that Muse Spark 1.1 is currently Meta's "most powerful model in AI intelligence and programming tasks." Compared to the first-generation Muse Spark released in April, which was only open to a few partners for API preview, the new version will be open to more developers through Meta's developer platform. Users can apply to join the waiting list and gradually obtain API access permission. However, Meta currently restricts API access to its own ecosystem and has not yet opened it to third-party model platforms such as OpenRouter. It is understood that Muse Spark 1.1 has significantly strengthened its code generation and AI intelligence capabilities. Wang stated that the reason the company prioritized improving programming capabilities is that code generation is the core foundation for AI intelligences to complete complex, multi-step tasks, helping to create AI assistants that can independently execute tasks. In terms of pricing, Meta has adopted a more competitive strategy. Wang stated that each new API account will receive a $20 free quota, with a fee of $1.25 per million tokens for input tokens and $4.25 per million tokens for output tokens, making it more attractive than similar products from companies like OpenAI and Anthropic, aiming to encourage developers to use it on a large scale. Earlier this week, Meta had just released the AI image generation model Muse Image, integrating it into core products such as Instagram, WhatsApp, Meta AI, and planning to apply it to advertising creative generation tools. The consecutive release of the Muse series of new products also reflects Meta's accelerated push towards commercializing AI in response to Wall Street's attention to its massive AI infrastructure investment returns. In recent years, Meta has been continuously expanding its AI capital expenditure, but compared to OpenAI, Anthropic, and Alphabet Inc. Class C (GOOGL.US), the company still lags behind in popular AI models and applications. At the same time, Meta has not yet established a mature cloud computing business, although the company has indicated plans to enter this market in the future. It is worth noting that Meta's AI strategy is also evolving. Previously, the company mainly relied on the open-source Llama series models to expand its developer ecosystem; now it is gradually shifting towards providing commercially available API services using its proprietary closed-source models to expand new sources of revenue. However, Wang emphasized that Meta has not abandoned its open-source strategy. The team is currently developing an open-source version of Muse Spark, but has not yet announced a release date. Additionally, he revealed that Meta is currently training a new generation AI model codenamed "Watermelon", which will further improve performance, but did not disclose a specific release date. He also mentioned that he has started using Muse Spark in scenarios such as health management, assisting decision-making through searching the web, reading academic papers, and integrating personal health data, believing this to be one of the important future applications of AI intelligences.