Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Multimodal & Grounded benchmark
PerceptionBench (Internal). Every model the catalog carries a published PerceptionBench value for, ranked by that value.
Moonshot AI's internal benchmark for atomic visual perception capabilities. 6 models in the ModelScale catalog have a published value.
Each point is a model with a published PerceptionBench value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
6 models with a published PerceptionBench value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Internal evaluation score |
|---|---|---|---|
| 1 | Moonshot AI | 68.3 | |
| 2 | TMInkling 2 Small | Thinking Machines Lab | 64.7 |
| 3 | Alibaba | 63.5 |
What the benchmark measures and how its value is published, from the source’s own record.
PerceptionBench (PerceptionBench (Internal)) is a multimodal & grounded benchmark published by Kimi K3: Open Frontier Intelligence. Moonshot AI's internal benchmark for atomic visual perception capabilities. Its evaluation set is described by the source as internal atomic visual-perception tasks. The source rates its difficulty as fine-grained visual perception.
Each model's result is published by Kimi K3: Open Frontier Intelligence as internal evaluation score. ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 6 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
All leaderboards · Multimodal & Grounded capability leaderboard
| 4 | Moonshot AI | 58.5 |
| 5 | TMInkling 2 Mini | Thinking Machines Lab | 57.5 |
| 6 | DSdots3-note Preview | Dots Studio | 53.4 |
Evidence key: Observed
Rows are ordered by the value Kimi K3: Open Frontier Intelligence published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.