Meta's Muse Spark Model: Big Improvements Reinvigorate Investors

Key Points

• Many investors saw Meta as having fallen behind in the AI race, but that perception looks to be changing.
• The company's latest Muse Spark model shows huge improvements over its Llama releases.
• After Meta invested over $14 billion to attract a top AI founder Alexandr Wang, the company moving in the right direction.

Investors seem to believe that Meta Platforms NASDAQ: META is making up ground in the artificial intelligence (AI) race following the disappointing results of its Llama models.

In April, the Magnificent Seven member released its latest AI model called Muse Spark. In the two days following that event, shares of Meta jumped more than 9%, easily surpassing the S&P 500’s 3% gain over the same period.

The market's bullish reaction largely came down to one thing: Muse Spark’s much-improved capabilities over Meta’s previous AI models. Combined with the company's demonstrated success in AI-enabled advertising, Muse Spark puts the company in a considerably better position going forward.

Muse Spark Shoots Meta Up AI Model Ranking Sites

Muse Spark significantly changes things for Meta, moving its top model from being somewhat of an afterthought to solidly competitive. Sites like Artificial Analysis help compare the intelligence of different AI models. The Artificial Analysis Intelligence Index currently gives Muse Spark a score of 52, making it a top-five ranking. Notably, it still sits behind top-tier models, including Google parent company Alphabet’s NASDAQ: GOOGL Gemini 3.1 Pro Preview, OpenAI’s ChatGPT 5.4, and Anthropic’s Claude Opus 4.6.

Still, Muse Spark isn’t too far off, with these higher-ranking models receiving scores between 53 and 57. Based on those numbers, Muse Spark may not be cutting edge, but the difference between Meta’s prior models is notable.

Meta’s most intelligent model prior to Muse Spark was LLama 4 Maverick, which according to Artificial Analysis received a score of just 18—nowhere close to the top models. The improvement on the site’s Agentic Index is even more striking, with Meta moving from a score of seven with Maverick to 62 with Muse Spark. However, the gap between Muse Spark and the top models on coding tasks remains somewhat wide.

LLM Arena is another commonly used AI model ranking site. On text-to-text tasks, it ranks Meta's Muse Spark as the second-best performing AI lab, with the model being the third-best model overall—above both Google and OpenAI. For context, Meta’s previous Llama models ranked outside of the top 100.

Ranking systems like these are not a perfect science, and day-to-day users often disagree with them. However, the jump that Meta appears to have made with Muse Spark is difficult to deny. The model’s release sets the company up to have a strong, general-purpose AI model to go along with an already impressive advertising business.

Muse Spark Creates New Levers Meta Can Pull

Muse Spark has several important implications for Meta. To start, it is the first model released by Meta Superintelligence Labs (MSL). The company formed MSL after paying $14.3 billion to buy 49% of Scale AI in 2025. In exchange for the investment, Alexandr Wang, Scale’s former CEO, agreed to become the leader of MSL.

With Wang at the helm, Meta’s models appear to have drastically improved in just nine months. This provides evidence that their massive bet on Wang is paying off. It also adds weight to the idea that Meta can continue to make even better models, with Wang still relatively early in his tenure.

Additionally, Meta plans to offer paid access to the model, which has the potential to create a significant, supplemental revenue stream. If Muse Spark is as good as the rankings suggest, and if Meta can further improve its models, users may actually have a reason to pay.

Meta will integrate Muse Spark into Facebook, Instagram, WhatsApp, Messenger, and its AI glasses. The goal of this is likely twofold. When consumers use Meta AI in these apps, Muse Spark's increased capabilities mean it can provide better answers, increasing engagement.

Furthermore, Meta can integrate Muse Spark into its advertising campaign tools. At the end of the day, advertisers are Meta’s real customers. Maximizing their return on ad spend (ROAS) is what keeps them spending more. With Muse Spark’s enhanced capabilities, Meta can improve existing tools and potentially roll out new ones. By creating a better experience for advertisers and improving ROAS, Meta can keep its nearly $200 billion advertising business chugging.

Despite Lowered Targets, Analysts Eye +30% Gains in Meta Stock

Notably, Meta received several price target updates after Muse Spark’s release on April 8. Interestingly, all of those tracked by MarketBeat moved lower.

However, these changes seem to be due to factors outside of Muse Spark, with analysts seeing the model itself as a positive. For example, Rosenblatt lowered its target despite calling Muse Spark "an encouraging development." The firm cited the Iran war as the reason for their decision, as rising gas prices could impact advertiser spend.

These updated targets average $845, still implying about 30% upside for the stock. This figure sits slightly above the MarketBeat consensus price target near $838.