JPMorgan recently upgraded Meta Platforms (META) from “Neutral” to “Overweight” and raised its price target on META stock from $640 to $820 after the company launched its Muse personal artificial intelligence (AI) agent. The AI agent reached third in the U.S. App Store shortly after launch, with Muse Spark appearing much better than the AI models Meta Platforms routinely used to release.
Muse is so good, in fact, that JPMorgan thinks the model can compete with leading AI models from OpenAI and Anthropic. Another model, called "Watermelon," is in the works from Meta and could further accelerate the company's progress against leading AI labs.
Why Muse Spark Is a Surprise
Meta Platforms has been making AI models for quite a while, but none have been as good as Muse Spark. Its Llama models barely made it into the benchmark charts of top AI models and only occasionally excelled on narrow tests. In fact, Llama 4 was delayed after it fell short of Meta's reasoning and math expectations last year, and the company ended up using an experimental version for a leaderboard instead of the model released to the public.
Muse Spark represents a ground-up reconstruction of that faltering program. Meta rebuilt its pretraining system across model architecture, optimization, and data curation. According to the company, Muse Spark 1.3 uses 20% fewer tool calls and 25% fewer tokens when compared to Muse Spark 1.2. That makes each dollar invested in training infrastructure considerably more productive.
The benchmark results finally place Meta within sight of the frontier. Muse Spark 1.3's highest-reasoning version scored 62 on the Artificial Analysis Intelligence Index. That puts it behind only two Anthropic models and roughly alongside the best systems from OpenAI and other leading laboratories. The model also led several demanding coding and long-context tests, including a 75.4% score on DeepSWE 1.1 and 98.5% on a long-context retrieval benchmark.
The AI Users Are Already There
OpenAI and Anthropic have to fight over users, but Meta Platforms doesn't really have to. Its Family of Apps segment includes Instagram, Facebook, Messenger, WhatsApp, and Threads. Almost all of these apps let you talk directly to Meta's AI models, and they have billions of users. Meta's AI now has some 1.5 billion users.
That said, the vast majority of these users are not paying users, and they are not going to pay anytime soon. Both OpenAI and Claude let people use far more capable AI models for $20-per-month subscriptions. Few people would likely pay for a Meta subscription right now, even though its AI capabilities have taken a leap.
What's exciting for Meta is that, if its capabilities keep accelerating and start getting closer to the frontier, the company could eventually convert a lot more users into paying customers. Advertising its AI models to billions would be a cakewalk, as they already have multiple Meta apps installed on their phones.
Could Muse Spark Replace Gemini?
It's hard not to notice how Meta Platforms is doing what Alphabet (GOOGL) is now failing to do. Last year, Google started “dominating” by releasing AI models that managed to compete in head-to-head benchmarks against frontier models from OpenAI and Anthropic. In fact, many users started shifting over to Google due to its more generous usage quotas and certain specializations. But that momentum has slowed sharply this year, and the gap between Google's models and those offered by OpenAI and Anthropic has only grown since Fable 5's release.
You'd be forgiven for thinking that AI was destined to be a game between the two major private companies after all. However, Meta could end up being the "Google" of 2026 or 2027 if its models keep getting better and start competing with the top models from Claude and OpenAI. That's certainly possible, especially when you consider the colossal amount of data that Meta has access to.
The average analyst agrees with JPMorgan and sees META stock rising higher. Overall, Meta Platforms stock has a consensus “Strong Buy” rating on Wall Street based on 54 analysts with coverage.
On the date of publication, Omor Ibne Ehsan did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. For more information please view the Barchart Disclosure Policy here.