multimodal benchmark
BLINK
BLINK: Multimodal Large Language Models Can See but Not Perceive. A benchmark for multimodal language models focusing on core visual perception abilities. Reformats 14 classic computer vision tasks into 3,807 multiple-choice questions paired with single or multiple images and visual prompting. Tasks include relative depth estimation, visual correspondence, forensics detection, multi-view reasoning, counting, object localization, and spatial reasoning that humans can solve 'within a blink'.
Updated Aug 7, 2026
Published models13
Registry coverage13
MetricScore
EvidenceB
BLINK leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | BY | 81.4% | 100.0% | 13 | C | |
| 2 | BY | 79.4% | 91.7% | 13 | C | |
| 3 | AC | 70.7% | 83.3% | 13 | C | |
| 4 | AC | 69.1% | 75.0% | 13 | C | |
| 5 | AC | 68.7% | 66.7% | 13 | C | |
| 6 | AC | 68.5% | 58.3% | 13 | C | |
| 7 | AC | 67.7% | 50.0% | 13 | C | |
| 8 | AC | 67.3% | 41.7% | 13 | C | |
| 9 | AC | 67.1% | 33.3% | 13 | C | |
| 10 | AC | 65.8% | 25.0% | 13 | C | |
| 11 | AC | 65.4% | 16.7% | 13 | C | |
| 12 | AC | 63.4% | 8.3% | 13 | C | |
| 13 | MI | 61.3% | 0.0% | 13 | C |
Score distribution
Top published rows on the benchmark's original scale.
BLINK highlights
The top published results on this benchmark's own scale.
What is BLINK?
Definition and scoring fields from the benchmark registry.
BLINK: Multimodal Large Language Models Can See but Not Perceive. A benchmark for multimodal language models focusing on core visual perception abilities. Reformats 14 classic computer vision tasks into 3,807 multiple-choice questions paired with single or multiple images and visual prompting. Tasks include relative depth estimation, visual correspondence, forensics detection, multi-view reasoning, counting, object localization, and spatial reasoning that humans can solve 'within a blink'.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- BLINK
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- blink|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about BLINK.
Which model scores highest on BLINK?
Seed 2.1 Pro is currently ranked first with 81.4%.
What does BLINK measure?
BLINK: Multimodal Large Language Models Can See but Not Perceive. A benchmark for multimodal language models focusing on core visual perception abilities. Reformats 14 classic computer vision tasks into 3,807 multiple-choice questions paired with single or multiple images and visual prompting. Tasks include relative depth estimation, visual correspondence, forensics detection, multi-view reasoning, counting, object localization, and spatial reasoning that humans can solve 'within a blink'.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
13 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.