Facebook parent Meta and Microsoft (MS) unveiled next-generation artificial intelligence (AI) models in quick succession, moving to strike back. The two big tech corporations had been seen as trailing OpenAI, Google and Anthropic in the AI market. Rather than trying to win with models that overwhelm rivals on performance, both companies aim to reclaim market leadership by linking state-of-the-art AI models to, and then advancing them through, their core businesses—social media (SNS) for Meta and cloud infrastructure for Microsoft.
After Meta unveiled its cutting-edge AI model "Muse Spark" on the 8th (local time), the company's stock closed up 6.5% from the previous day. While the performance of "Muse Spark" does not surpass rival models from OpenAI, Google or Anthropic, investors are focusing on the model's significant scalability and ease of monetization once it is applied to Meta's core platforms, which have billions of users.
Meta said, "Muse Spark is a model specifically designed for Meta's products," adding, "Meta AI, which runs on Muse Spark, will become faster and smarter, and in the future we will also roll out new features on Instagram, Facebook and Threads that leverage user-shared content and recommendation information."
Meta said "Muse Spark" shows strength in reasoning through complex questions in science, math and health, but its coding capability is relatively weaker than rival models. According to benchmark scores Meta released that day, "Muse Spark" delivered results on par with top-tier models such as OpenAI's "GPT-5.4," Google's "Gemini 3.1 Pro," and Anthropic's "Claude Opus 4.6" in some areas, including medical reasoning.
Although Meta's latest model is not the industry's best, it is sufficiently high-performing, and expectations that integrating it with the company's major SNS platforms will create strong synergy lifted market sentiment. Facebook, Instagram and WhatsApp—the three apps driving Meta's results—each have more than 3 billion users, making it advantageous for broad application of AI to drive diffusion and commercialization. In fact, Meta is working to introduce a paid subscription model to the three apps within the year, and paying users will be able to use a variety of AI features tailored to each platform's characteristics.
Wall Street investment banks set Meta's investment rating at "outperform" and projected a target price of $862 per share, about 40% higher than the current $612. Investment bank Mizuho said, "If Meta (by integrating Muse Spark) boosts user engagement around shopping and search functions and strengthens ad targeting, it could unlock new revenue generation opportunities." Bank of America (BofA) gave high marks to Meta for releasing "Muse Spark" earlier than expected, saying, "This early release has partially eased market uncertainty surrounding Meta's AI strategy."
Chief Executive Officer (CEO) Mark Zuckerberg last year poured astronomical funds into establishing the AI research unit "Meta Superintelligence Lab (MSL)." After Meta's existing AI model "Llama4" was criticized for lagging behind American rivals such as OpenAI and Google, as well as China's DeepSeek, the company moved to overhaul its AI organization by recruiting top talent across the board and acquiring AI startup Scale AI for $14.8 billion (about 22 trillion won), installing former CEO Alexander Wang as head of the superintelligence lab. At the time, some in the market suggested that results might fall short of expectations compared with Meta's aggressive AI investments, but concerns were allayed when the superintelligence lab, led by Chief AI Officer (CAIO) Alexander Wang, released a high-performance model in nine months.
MS, which is moving to end its partnership with OpenAI, also recently unveiled AI models specialized in voice transcription (dictation) and image and voice generation. The plan is to reduce reliance on OpenAI, with which it had maintained a cooperative relationship, and pursue self-sufficiency with in-house AI models.
On the 3rd, MS first unveiled three models in its MAI (MS AI) lineup. The total of three—voice transcription model "MAI-Transcribe-1," voice generation model "MAI-Voice-1," and image generation model "MAI-Image-2"—are all specialized for specific tasks rather than general-purpose models.
The company plans to focus on expense reduction and improved work efficiency through a vertically integrated strategy that connects the MAI lineup across its own cloud infrastructure Azure, its workplace chat tool Teams, PowerPoint and other in-house software and platforms. Its goal is to secure by next year a cutting-edge large AI model that handles text, images and voice, and to build a self-reliant AI system.
Mustafa Suleyman, MS AI CEO, said, "MAI targets higher performance, faster speed and lower expense than competitors," adding, "We are rapidly applying MAI's advanced models to MS's consumer and corporate products."