{"slug": "meta-launches-muse-spark-1-1-api-its-first-paid-ai-model-to-challenge-anthropic", "title": "Meta Launches Muse Spark 1.1 API, Its First Paid AI Model, to Challenge Anthropic and OpenAI", "summary": "Meta launched Muse Spark 1.1, its first paid AI model, and a commercial API to compete with Anthropic and OpenAI in the AI coding market. The closed-source model, priced at $1.25 per million input tokens and $4.25 per million output tokens, marks a strategic shift from Meta's open-source Llama tradition. Early API partners include Replit, Cline, and Box, with a public preview available to U.S. developers.", "body_md": "# Meta Launches Muse Spark 1.1 API, Its First Paid AI Model, to Challenge Anthropic and OpenAI\n\n- Meta launched Muse Spark 1.1 and a paid Meta Model API, its first commercial AI model offering, priced at $1.25 per million input tokens and $4.25 per million output tokens\n[[1]](https://www.cnbc.com/2026/07/09/meta-jumps-into-ai-coding-market-to-chase-anthropic-and-openai.html) - The closed-source model claims to rival OpenAI's GPT-5.5 and Anthropic's Opus-4.8 on agentic benchmarks, with a one-million-token context window\n[[2]](https://x.com/alexandr_wang/status/2075218936266998230) - Early API partners include Replit, Cline, and Box, with the public preview available to U.S. developers\n[[3]](https://www.testingcatalog.com/meta-debuts-muse-spark-1-1-model-and-opens-api-for-developers/) - The release marks a sharp strategic shift from Meta's open-source Llama tradition, driven by Chief AI Officer Alexandr Wang, who joined from Scale AI in 2025\n[[4]](https://letsdatascience.com/blog/meta-muse-spark-alexandr-wang-closed-model) - Meta shares traded flat at roughly $603 on the day of the announcement, with the stock down about 9% year-to-date\n[[5]](https://financialmodelingprep.com/)\n\nMeta Platforms on Wednesday released Muse Spark 1.1, an upgraded AI model purpose-built for coding and agentic tasks, alongside the company's first-ever paid developer API. The launch puts Meta in direct competition with Anthropic and OpenAI in the fast-growing market for AI-powered software development tools [1].\n\nThe model, developed by Meta Superintelligence Labs under Chief AI Officer Alexandr Wang, is available through a new Meta Model API at $1.25 per million input tokens and $4.25 per million output tokens — pricing that sits below Anthropic's flagship Claude Sonnet 4.6 but above entry-level offerings from both rivals [[1]](https://www.cnbc.com/2026/07/09/meta-jumps-into-ai-coding-market-to-chase-anthropic-and-openai.html) [3].\n\nMuse Spark 1.1 represents a significant departure for Meta, which built its AI reputation on open-source Llama models distributed under permissive licenses. The new model is proprietary with closed weights, accessible only through Meta's apps or the paid API — a philosophical reversal that reflects Wang's mandate to turn Meta's AI research into a revenue-generating business [4].\n\n## The Model\n\nMuse Spark 1.1 is a multimodal reasoning model supporting a one-million-token context window — large enough to process entire codebases in a single session. Meta describes it as capable of orchestrating multi-agent systems, functioning as both a primary agent and a subagent for complex workflows [2].\n\nThe model supports tool calling, function calling, computer-use workflows, and parallel subagent coordination. In coding tasks, Meta says it handles bug diagnosis, feature implementation in enterprise systems, and large-scale code migrations [3].\n\nWang said on social media that Muse Spark 1.1 'rivals GPT-5.5 and Opus-4.8' across agentic evaluation benchmarks, and claimed it achieved top positions on MedScribe, TaxEval, and Harvey's Legal Agent Bench while being '10x cheaper and twice as fast' [2].\n\nThe original Muse Spark 1.0, released in April, had notable coding weaknesses — scoring 59.0 on coding benchmarks versus 80.8 for Anthropic's Claude and 75.1 for OpenAI's GPT-5.4, according to the Artificial Analysis Intelligence Index [4]. The 1.1 upgrade is specifically targeted at closing that gap.\n\n## Pricing and Availability\n\nThe Meta Model API is available in public preview to U.S.-based developers, with $20 in free credits for new users. The API is OpenAI-compatible, lowering the barrier for developers to switch from competing platforms [3].\n\nAt $1.25 per million input tokens and $4.25 per million output tokens, Meta is positioning Muse Spark 1.1 as a cost-competitive alternative in the enterprise tier. The pricing undercuts Anthropic's Claude Sonnet 4.6 but sits above budget options like OpenAI's GPT-5 mini and Anthropic's Claude Haiku 4.5 [1].\n\nThe model is also available in 'Thinking' mode within the Meta AI consumer app and on meta.ai. Meta said Muse Spark 1.1 will replace existing Llama models powering chatbots across WhatsApp, Instagram, Facebook, and its Ray-Ban smart glasses [1].\n\n## Early Partners and Enterprise Push\n\nMeta named Replit, Cline, and Box as early API partners. Replit CEO Amjad Masad highlighted the million-token context window and OpenAI-compatible API format. Cline CEO Saoud Rizwan pointed to the model's tool usage capabilities and pricing advantages for scaled coding workloads [3].\n\nBox VP Yashodha Bhavnani cited competitive enterprise capabilities as a reason for the partnership [3].\n\nThe partner lineup signals Meta's ambition to compete in the agentic coding segment that Anthropic has dominated with Claude, which became the default model for AI-assisted coding tools in 2025 and 2026.\n\n## The Strategic Pivot\n\nThe closed-source approach is a striking reversal for a company that spent years evangelizing open AI development. Meta previously distributed Llama models freely, arguing that open-source AI benefited the ecosystem and countered the dominance of closed-model competitors [4].\n\nWang, the 28-year-old former co-founder and CEO of Scale AI, was recruited in June 2025 when Meta acquired a 49% nonvoting stake in Scale for $14.3 billion. He was appointed to the newly created Chief AI Officer role with a mandate to overhaul Meta's AI strategy after the underperformance of the Llama 4 model series [4].\n\nUnder Wang, Meta Superintelligence Labs debuted the original Muse Spark in April and released Muse Image, its first image-generation model, earlier this week [1]. The rapid release cadence suggests Wang is moving aggressively to establish Meta as a credible alternative to Anthropic and OpenAI in the enterprise AI market.\n\n## Market Context\n\nMeta shares traded roughly flat at $602.86 on Wednesday, giving the company a market capitalization of approximately $1.53 trillion. The stock is down about 9% year-to-date and 18% from its 52-week high of $796.25 [5].\n\nThe AI coding tools market has become one of the highest-growth segments in enterprise software. Anthropic's Claude has become the go-to model for developer tools, while OpenAI has expanded its own coding capabilities with GPT-5.5. Meta's entry with a paid API adds a third major competitor with the distribution advantage of billions of existing users across its family of apps.\n\nMeta's safety evaluation under its Advanced AI Scaling Framework found that unmitigated versions of Muse Spark 1.1 reached high-risk thresholds in chemical, biological, and cybersecurity domains, but multi-layered safeguards reduced residual risk to moderate or lower levels. The mitigated model showed lower jailbreak success rates compared to Muse Spark 1.0 [3].\n\n## Further sources\n\n[[1] CNBC: Meta jumps into AI coding market in effort to chase Anthropic and OpenAI ↗](https://www.cnbc.com/2026/07/09/meta-jumps-into-ai-coding-market-to-chase-anthropic-and-openai.html)\n\n[[2] Alexandr Wang on X announcing Muse Spark 1.1 benchmarks and availability ↗](https://x.com/alexandr_wang/status/2075218936266998230)\n\n[[3] TestingCatalog: Meta debuts Muse Spark 1.1 model and opens API for developers ↗](https://www.testingcatalog.com/meta-debuts-muse-spark-1-1-model-and-opens-api-for-developers/)\n\n[[4] Let's Data Science: Meta Muse Spark: Alexandr Wang Delivers First Closed-Source… ↗](https://letsdatascience.com/blog/meta-muse-spark-alexandr-wang-closed-model)\n\n[[5] Meta Platforms real-time stock quote via Financial Modeling Prep ↗](https://financialmodelingprep.com/)\n\nThe stories that matter, in one email. Free — unsubscribe anytime.", "url": "https://wpnews.pro/news/meta-launches-muse-spark-1-1-api-its-first-paid-ai-model-to-challenge-anthropic", "canonical_source": "https://mlq.ai/news/meta-launches-muse-spark-11-api-its-first-paid-ai-model-to-challenge-anthropic-and-openai/", "published_at": "2026-07-09 19:35:00+00:00", "updated_at": "2026-07-12 17:36:39.915101+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products", "ai-startups", "ai-policy"], "entities": ["Meta", "Muse Spark 1.1", "Anthropic", "OpenAI", "Alexandr Wang", "Replit", "Cline", "Box"], "alternates": {"html": "https://wpnews.pro/news/meta-launches-muse-spark-1-1-api-its-first-paid-ai-model-to-challenge-anthropic", "markdown": "https://wpnews.pro/news/meta-launches-muse-spark-1-1-api-its-first-paid-ai-model-to-challenge-anthropic.md", "text": "https://wpnews.pro/news/meta-launches-muse-spark-1-1-api-its-first-paid-ai-model-to-challenge-anthropic.txt", "jsonld": "https://wpnews.pro/news/meta-launches-muse-spark-1-1-api-its-first-paid-ai-model-to-challenge-anthropic.jsonld"}}