Muse Wonder lands in Cursor with 1M context

The configurable effort levels are more compelling than the headline context number: developers can tune Muse Wonder per task instead of paying frontier-model costs everywhere.

• –Six reasoning settings make it practical to balance latency, quality, and spend
• –The 1M-token Max Mode should help with large repositories and long tool-use chains
• –Cursor access brings Meta’s newest model into a direct comparison with leading coding models
• –At $4 per million input tokens and $20 per million output tokens, heavy reasoning runs will be expensive
• –Real-world coding evaluations—not context size alone—will determine whether it actually outperforms Cursor’s alternatives