Meta’s Muse Spark 1.1 tops GLM-5.2 in coding benchmarks
Meta’s Muse Spark 1.1 just leapfrogged GLM-5.2 in coding prowess while keeping the price tag slightly lower, delivering a sharper, cheaper alternative for developers. On the Artificial Analysis Intelligence Index, Muse Spark 1.1 now sits at 51, an eight-point jump in three months. In coding benchmarks it tops GLM-5.2 with a score of 71.3, all for $0.26 per task—about a tenth of a cent less than its rival.
A measurable leap in performance
The update marks a clear uptick in Muse Spark 1.1’s capabilities. The model’s hallucination rate has also fallen sharply, from 73 percent down to 38 percent, reducing the need for costly post-generation verification. These figures suggest Meta is steadily closing the gap with leading open-weight models in both accuracy and efficiency.
Why cost and reliability matter
For teams balancing budgets and quality, Muse Spark 1.1’s lower price and reduced hallucination rate translate to faster iteration cycles and fewer manual fixes. The model’s improved index score further signals broader gains in reasoning and general intelligence, which could influence everything from enterprise chatbots to autonomous coding assistants.
Why it matters
Muse Spark 1.1 isn’t just another incremental release; it’s a concrete step toward more capable and affordable AI tools for developers. By combining higher coding scores with lower costs and hallucination rates, Meta is making a model that’s both technically stronger and economically viable. That balance could push more teams to adopt open-weight alternatives over pricier proprietary options.
Source: The Decoder. AI-assisted editorial synthesis — TechnoExpress.