Ten papers covering the canonical multimodal benchmarks. MMBench's circular eval is the position-bias control to copy. MMMU and MMMU-Pro are the standard college-level VQA pair, with MMMU-Pro tightening the loop by stripping text-only-solvable questions. BLINK is the result to cite when someone says perception is solved, and BabyVision (January 2026) pushes the point further: its 388 items test visual primitives a young child handles, and Gemini3-Pro-Preview scored 49.7 against 94.1 for adults. Video-MME and Vibe-Eval push into video and harder prompts. Video-MME-v2 (April 2026) is the video version to run now: 800 videos, over 80% published in 2025 or later to limit contamination, four related questions per video scored as a group, so one lucky answer earns little. The best model scored 49.4 there against 90.7 for human experts, and the same model's plain per-question accuracy was 66.1%, which is the gap per-question scoring hides. ChartQA remains the chart-reasoning baseline worth re-verifying before publication.