Meta Unveils Muse Spark to Challenge OpenAI and Google

Meta has unveiled a new artificial intelligence model called Muse Spark, marking its first major AI release since the company reorganized its strategy and made a multibillion dollar push to catch up with rivals like OpenAI, Anthropic, and Google. The new model is being presented as an important turning point for Meta after the disappointing rollout of its earlier Llama 4 models, which failed to gain strong traction with developers. Now, with Alexandr Wang leading the company’s new Meta Superintelligence Labs after Meta’s $14.3 billion investment in Scale AI, the company is betting that Muse Spark can help restore momentum in one of the most competitive areas in tech.

Muse Spark, previously known internally as Avocado, was developed over the past nine months and is the first model in Meta’s new Muse series. Meta says the model is small and fast by design, but still capable of handling more advanced questions in areas such as science, math, and health. Rather than positioning it as the strongest model on the market, the company is emphasizing efficiency, competitive performance, and the speed of its rebuilt AI development cycle. Meta says improvements in training techniques and a rebuilt technology stack allowed it to create a smaller model with capabilities comparable to an older midsize Llama 4 system while using far less computing power. Bloomberg also reported that Muse Spark was trained using several third party open source models, including Qwen as well as models from OpenAI and Google, showing how aggressively Meta is trying to accelerate development.

The release also reflects a broader strategic shift inside Meta. While the company built much of its recent AI reputation around open source Llama models, Muse Spark is proprietary and closed, meaning its design and code will not be publicly released. That change appears to align with Wang’s influence and Meta’s desire to create a more commercially flexible AI business. The company is already testing a new revenue stream by giving select partners access to Muse Spark through a private API preview, with plans to eventually offer paid access more widely. Meta is also reportedly considering subscription fees for its Meta AI chatbot in the future, even though the assistant remains free for now.

At the same time, Meta is spending at an enormous scale to remain competitive. The company expects AI related capital expenditures in 2026 to land between $115 billion and $135 billion, nearly double the year before. Zuckerberg has also spent billions on AI talent and infrastructure such as data centers as he tries to close the gap with competitors that have surged ahead in the AI race. Meta has acknowledged that Muse Spark is not yet as capable as some top offerings from OpenAI, Anthropic, or Google, especially in areas like coding and longer horizon agentic systems, but the company describes it as an early signal of where its new AI lab is heading, with several larger models already in development.

Muse Spark is also central to Meta’s plan to weave AI more deeply across its consumer products. The model already powers the Meta AI assistant in the standalone app and desktop website, and it is expected to roll out in the coming weeks across Facebook, Instagram, WhatsApp, Messenger, and Ray Ban Meta smart glasses. Users will be able to switch between different modes depending on how complex their task is, from quick responses to more sophisticated analysis. Meta says a Contemplating mode will gradually roll out for advanced prompts, using multiple AI agents to reason in parallel and compete with the higher end reasoning modes offered by rival frontier models. The assistant is also being built to handle tasks such as analyzing legal documents, identifying nutritional information from grocery photos, and supporting broader health related analysis.

Beyond answering questions, Muse Spark is also being positioned as a tool for shopping and discovery. Meta says the updated assistant will help people find clothes, furniture, and other products through a shopping mode that draws on creator content, brand storytelling, and community activity already happening across its apps. The model is also expected to power the company’s Vibes AI video feature over time. In many ways, Muse Spark is not just a new model launch but an attempt to reset Meta’s broader AI business. After setbacks with earlier releases, the company is trying to prove that it can still compete by building faster, spending bigger, and turning AI into both a product feature and a future revenue engine.

Works Cited

Bihler, Dennis. “Meta Unveils Muse Spark AI Model to Regain Ground in Race with OpenAI and Google.” Ynetnews, 8 Apr. 2026, www.ynetnews.com/tech-and-digital/article/bycs9u4nzl. 

Griffin, Riley, and Kurt Wagner. “Meta Debuts First AI Model From New Superintelligence Group.” Bloomberg, 8 Apr. 2026, www.bloomberg.com/news/articles/2026-04-08/meta-debuts-first-ai-model-from-prized-superintelligence-group. 

Vanian, Jonathan. “Meta Debuts New AI Model, Attempting to Catch Google, OpenAI After Spending Billions.” CNBC, 8 Apr. 2026, www.cnbc.com/2026/04/08/meta-debuts-first-major-ai-model-since-14-billion-deal-to-bring-in-alexandr-wang.html. 

Meyer, Anton. “Meta Ray-Ban Display: You Can Control These Glasses Using Gestures.” nextpit, 19 Sept. 2025, www.nextpit.com/news/meta-ray-ban-display-control-your-glasses-using-gestures. 

Mattson, Jennifer. “Who Is Alexandr Wang? Scale AI’s Young Billionaire CEO Will Head Meta’s New ‘Superintelligence’ Lab.” Fast Company, 10 June 2025, www.fastcompany.com/91349929/alexandr-wang-scale-ai-ceo-to-head-meta-superintelligence-lab-what-to-know.