Meta Muse Spark 1.1 Scores 71 on Coding Benchmark

Meta's Muse Spark 1.1 model has achieved a score of 71 on an independent coding benchmark. This new data comes from Artificial Analysis, following the model's API launch on July 9th. What's interesting is the cost efficiency. Muse Spark 1.1 costs about $0.26 per task, significantly less than rivals like GLM-5.2 at $0.37 and GPT-5.4 at $0.89, even though all three scored 51 on the Intelligence Index. This is due to Muse Spark 1.1 generating fewer output tokens. The model also shows an 8-point gain on the Intelligence Index, moving from 43 to 51. These improvements are concentrated in scientific reasoning, coding, and knowledge. The coding sub-score specifically rose by 12 points, from 59 to 71. Its performance on the SciCode scientific reasoning benchmark reached 58 percent, ranking third among evaluated models. The context window for Muse Spark 1.1 also expanded to 1 million tokens, a notable change for coding tasks. The output speed is approximately 114 tokens per second, which is faster than many comparable reasoning models. This means developers could see improved performance at a lower cost.