reasoning benchmark
CountBench
CountBench evaluates object counting capabilities in visual understanding.
Updated Aug 7, 2026
Published models6
Registry coverage6
MetricScore
EvidenceB
CountBench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 0.978 points | 100.0% | 6 | C | |
| 2 | AC | 0.978 points | 80.0% | 6 | C | |
| 3 | AC | 0.978 points | 60.0% | 6 | C | |
| 4 | AC | 0.976 points | 40.0% | 6 | C | |
| 5 | AC | 0.97 points | 20.0% | 6 | C | |
| 6 | AC | 0.937 points | 0.0% | 6 | C |
Score distribution
Top published rows on the benchmark's original scale.
CountBench highlights
The top published results on this benchmark's own scale.
What is CountBench?
Definition and scoring fields from the benchmark registry.
CountBench evaluates object counting capabilities in visual understanding.
Scores are shown in points. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- CountBench
- Modality
- image
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- countbench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about CountBench.
Which model scores highest on CountBench?
Qwen3.5-27B is currently ranked first with 0.978 points.
What does CountBench measure?
CountBench evaluates object counting capabilities in visual understanding.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
6 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.