Meta Developed Proprietary AI Model Codenamed Watermelon

The company has shifted its development strategy away from open-weight models toward a proprietary approach.

Updated on Sept. 23, 2026 in Artificial Intelligence

Isometric editorial illustration of a metallic server rack, representing technical infrastructure development. — Meta announced a shift in its artificial intelligence strategy, prioritizing the development of a new proprietary model codenamed Watermelon for its hardware ecosystem. AI Illustration. Upload story photo >

Live Poll

Do you trust proprietary AI models from big tech companies more than open-source alternatives?

Meta Chief AI Officer Alexandr Wang announced that the company is preparing to release a new proprietary AI model codenamed Watermelon. The model is reportedly designed to reach performance parity with OpenAI's GPT-5.5.

Why it matters

The company adjusted its strategy to focus on deep integration within its apps and hardware ecosystem following widespread criticism regarding the release of Llama 4.

The Watermelon model utilizes significantly increased computational resources to achieve its performance goals. This marks a departure from Meta's previous reliance on open-weight models like the Llama 4 series.

The players

Alexandr Wang

He serves as the Chief AI Officer at Meta Platforms and is leading the development of the company's new AI strategy.

Meta Platforms

This technology conglomerate owns social media applications and hardware ecosystems currently being integrated with advanced AI models.

The details

Meta is prioritizing proprietary systems that are tightly woven into its existing app and hardware infrastructure. The model reportedly targets competitive performance metrics against advanced industry benchmarks like OpenAI's GPT-5.5.

Meta released Llama 4 in April 2025.
Alexandr Wang joined the Meta team in June 2025.
The Muse Spark model was launched in April 2026.
Meta released Muse Spark version 1.3 in September 2026.

The Tech Race

This transition to proprietary systems reflects a broader industry arms race, as firms increasingly guard their underlying models to secure competitive advantages. It represents a pivot away from the open-weight development model that defined Meta's previous AI output.

Users can expect AI capabilities to become more deeply integrated into their daily interactions with Meta's apps and hardware. These proprietary systems are designed to enhance functionality across the ecosystem, though they may limit public access to the underlying code.

The takeaway

Meta is betting that proprietary AI integration will drive future growth and differentiation for its product ecosystem. This shift marks a significant change in how the company approaches the open-source community versus its commercial objectives.