Meta has released an AI agent for programming. Is this a competitor to OpenAI and Anthropic?

Meta's Muse Code offers two pricing plans: the cheaper one requires users to help the company develop the agent / Photo: Rokas Tenys / Shutterstock.com
Meta Platforms wants to carve out a niche in one of Silicon Valley’s hottest markets—the market for AI agents for programming, according to The Wall Street Journal. The owner of Facebook unveiled a new agent, Muse Code, on Wednesday, positioning it as a cheaper alternative to competitors’ solutions.
“We believe this could be an exceptionally good option for many workflows and use cases, especially from a cost perspective,” Meta’s head of AI, Alexander Wang, told the WSJ. According to him, some Meta employees are already using the agent internally, and he expects the scale of internal adoption to grow.
Meta’s Muse Code offers two pricing tiers: the first costs the same as the all-purpose Muse Spark model, and the second is more than ten times cheaper. To access the budget plan priced at $0.2 per 1 million outgoing tokens, users must agree to provide feedback that will help improve the agent. Tokens are the basic unit of computation in artificial intelligence systems.
Similar solutions—Claude Code from Anthropic and Codex from OpenAI—are included in plans costing around $20 per month and have computational limits. Heavier usage will cost more, and once the limit is exceeded, users switch to pay-as-you-go billing. The cost of 1 million outgoing tokens for comparable OpenAI models ranges from $12 to $30, while for Anthropic models, it ranges from $10 to $25.
The price of Meta’s new agent is roughly on par with that of Chinese models such as DeepSeek, where 1 million output tokens can cost as little as $0.18, according to the WSJ. OpenAI also slashed prices for simpler models on July 30. For example, the cost of GPT-5.6 Luna fell from $6 to $1.20 per 1 million tokens.
Last summer, Meta revamped its AI strategy and created a new division, Meta Superintelligence Labs. It is led by Alexander Wang, co-founder of Scale AI, a lab that develops AI datasets. He began aggressively recruiting specialists, offering multi-million compensation packages to employees at leading AI labs, and set out to develop Muse Spark—Meta’s new AI model, released in April. On Wednesday, Meta also unveiled Muse Spark 1.2.
Meta shares rose 0.1% to $588.8 during regular trading on Wednesday, August 5. In after-hours trading, the stock gained another 0.6%.
This article was AI-translated and verified by a human editor





