Cursor ships Grok 4.5 model outperforming GPT-5.5 at half the cost

Cursor released Grok 4.5 as a model option, with benchmarks showing it outperforming GPT-5.5 at approximately 50% of the cost. The release included CursorBench results comparing performance across coding tasks. Multiple posts report Grok 4.5 appearing as the default model option in Cursor, signaling a shift in model selection among developers using the coding agent platform.