Meta launched this Wednesday the Muse Code, an artificial intelligence tool (IA) capable of programming autonomously, thus entering the most lucrative segment of the generating AI, dominated by Anthropic and OpenAI.
© Getty Images
06/08/2026
by
Lusa
A programming assistance is today the main market where AI generates concrete recipes, beyond language models such as ChatGPT or Gemini.
Microsoft was the pioneer with GitHub Copilot, before the emergence of a new generation of so-called “agency” tools, capable of chaining several complex tasks alone: Claude Code at Anthropic, Codex at OpenAI or Cursor, whose acquisition by SpaceX, the Elon Musk group, is in the process of completion.
Meta's leader, Mark Zuckerberg, explained on his social network Threads, that Muse Code is not a simple computer code generator and broker, but that it plans to carry out a project, divides it into tasks, writes the code and checks the result, and can delegate tasks to several “subagents” working in parallel.
The tool, in test version, will be billed by use and not by subscription, with rates lower than the competition, in a strategy taken to win the market.
Meta even proposes, by default, a tariff about ten times lower than those users who agree to share their user data to improve their products.
The revenues of the social media giant, which owns Facebook, Instagram, and WhatsApp, depend almost on the whole of the advertising, thus trying to rent directly their AI models under pressure from investors concerned with the size of their spending.
The Muse Code is based on the latest version of Muse Spark, the model presented in April, the first fruit of the complete reformulation of the AI in the Meta, whose previous models were considered to be at a disadvantage to competition.
In the middle of 2025, Mark Zuckerberg disbursed more than 14 billion dollars (12.1 billion euros) to acquire 49% of Scale AI and put its leader, Alexandr Wang, ahead of a new entity, Meta Superintelligence Labs, while hiring the weight of gold paintings from OpenAI, Anthropic and Google.
The multimillionaire also broke at that time with the tradition of the house of open models of free and free access, defended by French researcher Yann LeCun, who left the group at the end of 2025 to create his own ‘startup’.
The artificial intelligence security (IA) controls that the United States Government intends to apply will target the closed models of OpenAI, Google and Anthropic, saving open Meta or DeepSeek models, reported this week ‘media’ Americans.
Closed AI models remain housed on their creator's servers, which maintain control over their use without revealing the programming. On the contrary, the user of an open AI model can download and modify it.