Meta has transitioned its smart glasses to use the new Muse Spark AI model, marking a significant upgrade over its previous Llama 4 system. This change narrows the gap between Meta’s offerings and those from leading competitors like OpenAI (GPT), Google DeepMind (Gemini), Anthropic (Claude), and xAI (Grok).

Announced in April, Muse Spark is the first model publicly released by Meta Superintelligence Labs (MSL), a group established last year with billions of dollars invested in top AI talent. According to Meta, Spark is part of its Muse series, which succeeds the Llama series. While earlier models like Llama were open-source, this new iteration will remain proprietary, though the company hopes future versions might be open-sourced.

The shift from Llama 4 was driven by performance issues. The Llama models had fallen significantly behind leading AI systems, prompting Mark Zuckerberg to lay off a large portion of the team and initiate a series of high-profile hires. Despite being smaller and faster than its predecessor, Muse Spark shows promise, rivaling larger models in certain benchmarks while using 10 times less computational resources.

Meta reports that Muse Spark matches the performance of Llama 4 Maverick, its previous best model, but it outperforms it by a significant margin when considering compute efficiency. This smaller size makes it ideal for real-time responses on both traditional platforms and smart glasses, where quick response times are crucial for user experience.

On the Artificial Intelligence Index, Muse Spark scores 52, compared to Gemini 3.1 Pro at 57, GPT-5.5 at 60, and Claude Opus 4.8 at 61. While not yet on par with top tier models, these scores indicate progress.

In May, Meta implemented Muse Spark across its Ray Ban and Oakley smart glasses except for the Meta Ray Ban Display model, which continues to use a custom version of Llama 4. The remaining models now offer faster, more efficient responses, enhancing the overall user experience on smart glasses.

The continued use of Llama 4 on some devices highlights the ongoing challenges in fully adopting new AI technologies across all platforms. However, Meta’s efforts indicate a narrowing gap with leading AI labs like Google’s Gemini. In May, Google released Gemini 3.5 Flash and is set to release Gemini 3.5 Pro soon. This suggests that the competition between these companies will continue to evolve rapidly.

For Meta, this transition underscores its long-term goal of integrating always on contextual AI into smart glasses before or at least on par with competitors like Google. The success of this initiative could be crucial for establishing Meta as a leading player in the smart glasses market.

Source: https://www.uploadvr.com/meta-muse-spark-ai-model-replaces-llama-on-smart-glasses/

Thinking about building an AI product?

Get in Touch