Skip to content
AI IntelligenceJul 11, 2026AI Intelligence
Article

Meta's Muse Spark 1.1 outperforms GLM-5.2 in coding benchmarks and costs less per task.

Its hallucination rate dropped from 73% to 38%. The model improved by eight points in three months, showcasing the intense competition among large language models.

Data Cube AI EditorialSource: The Decoder
01

Source Brief

Meta's Muse Spark 1.1 outperforms GLM-5.2 in coding benchmarks and costs less per task. Its hallucination rate dropped from 73% to 38%. The model improved by eight points in three months, showcasing the intense competition among large language models.

02