I tested Muse Spark 1.1, a new close-to-frontier LLM from Meta Superintelligence Labs, on my Rust program. Given that I have used small local models to maintain my Rust program, I felt Muse Spark should have no trouble. I had a fairly simple task I wanted done.
I set up the model in OpenCode with OpenRouter. I had to verify that I am indeed at least 18 years of age before it would allow me to continue. I then issued my prompt to Muse Spark and it set about its work.
Muse Spark successfully completed my task, but:
• It took 2 prompts for a relatively simple fix—I am unsure whether my prompt was not clear, or if Muse Spark was trained to do the minimum possible work to fix something.
• The changes it made were correct and matched what I wanted (after the second prompt).
• Using Muse Spark was easier than doing the code myself, and it cost 33 cents.
This might not have been a great test, as I had an exact fix in mind, but Muse Spark performed well. It seems to be terse and doesn't chat endlessly about nonsense, which I appreciated. It finished with the statement "No match or if" instead of a 500-word essay.