What happened

Meta's Muse Spark 1.1 scored 51 on the Artificial Analysis Intelligence Index, according to the supplied research record. That is an increase of eight points over three months.

On the cited coding measure, Muse Spark 1.1 scored 71.3 and reportedly edged GLM-5.2. The record also says its cost was $0.26 per task, making the comparison notable on both coding performance and reported price.

Why it matters

The update adds to a broader, still limited signal that models positioned on efficiency may be closing selected gaps with competitors. For teams assessing coding tools, a narrow benchmark lead paired with a lower reported cost could justify a direct comparison with their current setup.

The supplied implication is practical rather than conclusive: test Muse Spark 1.1 against an existing stack using the team's own tasks. Index scores can be useful screening data, but they do not establish performance across every coding workflow.

The important limit

This is a thin, single-source research record with medium confidence, and the reported edge concerns one coding metric. The update is described as incremental, so it should not be read as proof of a broad or durable competitive lead.

The record also reports a decline in Muse Spark 1.1's hallucination rate from 73% to 38%. That is a substantial reported change, but the supplied evidence does not provide methodology or independent confirmation.

What to watch

Watch for a head-to-head evaluation on real coding tasks, including whether the reported $0.26 per-task cost and 71.3 coding score hold in comparable use.

Receipts

Upstream references

Digest dated 2026-07-12 · upstream model claude-sonnet-4-6. Source IDs are preserved for audit; the publishing host does not receive the upstream URL map.

  1. 1
    8b523f69272f43b08bfdc8332c146aea67fafc58Reference from the upstream research server

This quick brief was generated by Terra from a dated upstream research digest. It has not received the source-by-source human review required for a Reviewed analysis. Material limit: The evidence is effectively single-source and does not include benchmark methodology, independent validation, or evidence that the narrow coding result generalizes beyond the cited metric.