Meta's Muse Spark 1.1 Beats GLM-5.2 in Coding

Meta just dropped an updated version of its Muse Spark model, and it’s making waves in the AI coding arena. Muse Spark 1.1 now outpaces GLM-5.2 on coding benchmarks, scoring 71.3 on the Coding Index versus GLM’s 68.8. It also ties with top contenders like GPT-5.4 on the Intelligence Index at 51. The model saw an impressive eight-point jump in just three months, driven by improvements in coding and agent-based tasks. Perhaps more importantly, Muse Spark 1.1 is cheaper: $0.26 per task compared to $0.37 for GLM-5.2 and $0.89 for GPT-5.4. Meta also slashed the hallucination rate from 73% to 38%—the model now often refuses to answer rather than guessing wrong—and quadrupled the context window to a million tokens. For now, it’s only available via Meta’s own API. Why it matters: This shows how rapidly open-model ecosystems are closing the gap with proprietary leaders, potentially reshaping the cost-performance calculus for developers.