Meta AI: New Muse Spark Model Rivals GPT-5.5 in Coding

Meta Platforms is preparing to release a new version of its Muse Spark artificial intelligence model soon. The company's Chief AI Officer, Alexandr Wang, announced this update. What's interesting is this new algorithm is reportedly competitive with GPT-5.5 on several AI benchmarks. Wang stated the upcoming model is significantly more adept at coding than the original Muse Spark. It also improves the performance of AI agents. The original Muse Spark scored 52.5% on the SWE-Bench Pro coding test, while GPT-5.5 achieved 58.6%. The new Meta model is expected to surpass Muse Spark's coding abilities. An X user asked Wang about matching Claude Opus 4.8's coding capabilities, and he replied, "pretty soon." Claude Opus 4.8 scored 69.2% on SWE-Bench Pro. This improved performance reportedly requires significantly more computing capacity than the previous Muse Spark. Meta's focus on coding capabilities suggests it may aim to make these models available to external developers. This could impact how developers interact with AI tools in the future.