Meta jumps into AI coding market to chase Anthropic and OpenAI

Source Domain: www.cnbc.com

Here’s a summary of the key points presented in the article from the Bloomberg conference concerning Meta’s AI initiatives:

Muse Spark 1.1 Release: Meta unveiled Muse Spark 1.1, an advanced AI model developed under the leadership of chief AI officer Alexandr Wang, positioning it as the company’s most advanced model for agentic and coding tasks.
Limited and Public Previews: Initially available to select partners through a private API, Muse Spark 1.1 is now undergoing a public preview with developers able to join a waitlist for access.
Focused Access: For now, access to the new model’s API is limited to Meta’s own properties rather than being available on third-party platforms like OpenRouter.
Aggressive Pricing: The new model offers competitive pricing, starting with $20 in free credits for new API accounts, followed by $1.25 per million tokens of input and $4.25 per million tokens of output.
Performance and Adoption Goals: Muse Spark 1.1 is designed to interact well with popular coding harnesses, aiming to enhance AI agents’ capabilities to perform multiple tasks.
Commitment to Open Source: While focusing on proprietary models, Meta has also developed an open-source variant of Muse Spark, although the timeline for its release remains undisclosed.
Future Models: Meta is currently developing a more powerful AI model, code-named Watermelon, although details regarding its launch remain undisclosed.
Application in Health: Wang is exploring how the latest AI model could assist with personal health tasks, such as web searches, reading academic papers, and accessing health-related data.