Researchers introduce IntReal, a benchmark addressing the underexplored task of multimodal interleaved generation where text and images are naturally interwoven. The framework comprises two components: IntReal-100K, a real-world dataset containing 100,000 interleaved samples spanning four major categories, and the IntReal Evaluation Suite, which assesses model performance across seven complementary dimensions using specialized expert models to enable coarse-to-fine diagnosis. The evaluation suite is validated through three key metrics: low inter-metric redundancy with a maximum pairwise absolute Spearman's ρ of 0.42, high agreement with human judgments at 83.3% accuracy, and strong cross-evaluator consistency with Spearman's ρ exceeding 0.9. Benchmarking four families of Unified Multimodal Models (UMMs) reveals distinct capability profiles, exposing specific strengths and weaknesses in interleaved generation. Post-training experiments demonstrate that fine-tuning on IntReal-100K improves interleaved generation performance, and notably, that optimizing unimodal generation capabilities independently can also enhance interleaved output. These results suggest that interleaved generation benefits from modality-specific optimization strategies rather than requiring joint text-image training at every stage, offering a more efficient pathway for advancing UMM capabilities.