Meta's Muse Spark 1.1 Tops GLM-5.2 on Coding Tasks

Meta pushed out an update to its Muse Spark model on Thursday, and the numbers put it ahead of Zhipu's GLM-5.2 on coding tasks for the first time. Muse Spark 1.1 posted a 71.3 on the Coding Index, tracked by Artificial Analysis, compared to 68.8 for GLM-5.2. That puts Meta's model within a tenth of a point of GPT-5.6 Luna, which leads that particular measure at 71.4.

The gain didn't happen overnight. Muse Spark has added eight points on the broader Intelligence Index in three months, with most of that improvement coming from coding and agent-based knowledge work, the kind of tasks that involve chaining tool calls and multi-step reasoning rather than answering a single question. On the overall Intelligence Index, Muse Spark 1.1 now sits at 51, putting it level with GLM-5.2, GPT-5.4, and GPT-5.6 Luna.

Price is where the comparison gets more interesting. Meta's model reportedly costs less per task than GLM-5.2, and the gap isn't small once you factor in how many tokens each one burns through to get an answer. That's before getting into what changed under the hood, and why Meta says the model now hallucinates far less than it used to.