Muse Spark 1.2 reported on Artificial Analysis's Intelligence Index vs. Cost per Task Pareto frontier at $0.40 per task
Meta's Muse Spark 1.2 reaches cost-efficiency Pareto frontier at $0.40 per task
The Pareto frontier placement means Muse Spark 1.2 offers a cost-intelligence tradeoff that no compared model in the index matches. The cybersecurity safety evaluation provides a public record of the model's autonomous offensive cyber capability, including its ceiling on hard, long-horizon challenges.
The full picture
Muse Spark 1.2 sits on Artificial Analysis's Intelligence Index vs. Cost per Task Pareto frontier at $0.40 per task, meaning no compared model is simultaneously cheaper and higher-scoring on that index. It reaches approximately Claude Opus 4.8-level intelligence at about one-fifth of Opus 4.8's $2.03 per-task cost, and scores 6 Intelligence Index points below Claude Opus 5 at roughly one-sixth of Opus 5's cost. The cost-per-task figure accounts for input, cache, reasoning, and answer tokens consumed across nine benchmark evaluations, not list price alone. The higher cost compared to Muse Spark 1.1 at $0.29 per task reflects heavier token usage on agentic tasks rather than a price change. Meta also published a safety and preparedness report showing Muse Spark solved 100% of easy CyScenarioBench offensive cyber challenges and 46% of medium ones, but only 8% of hard ones. The evaluation used 37 Irregular challenges across atomic and end-to-end categories, with a challenge counted as solved if at least one of 20 runs succeeds.
How it developed
Sources
Want this in your inbox?
I send a short email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free