Meta Upgrades Muse Spark
Meta is taking on OpenAI and Anthropic with a massive upgrade to its flagship Muse Spark model. 💻 Here is what we know about the upcoming coding powerhouse.
What the video says
Meta is gearing up to launch an updated version of its flagship Muse Spark artificial intelligence model to challenge OpenAI and Anthropic in coding. The new algorithm reportedly matches OpenAI's GPT-5.5 across several closely followed benchmarks, though this performance comes at a cost.
According to Business Insider, the improved output quality of this upcoming model requires an order of magnitude more computing capacity. The original Muse Spark scored 52.5% on the SWE Bench Pro coding evaluation, while GPT-5.5 reached 58.6%.
Anthropic's Claude Opus 4.8 leads that test at 69.2%, but fell behind GPT-5.5 on Terminal Bench 2.0. Meta's Chief AI Officer Alexander Wang confirmed on X that a model matching Claude's coding capabilities is coming pretty soon.
Wang noted the new model is significantly more adept at coding and powering AI agents than the original Muse Spark. This hardware-heavy push comes as Meta also debates launching an AI infrastructure service to offer raw computing capacity or hosted models.
