LLM Stats (PerceptionBench): leaderboard
PerceptionBench is Moonshot AI's internal benchmark of atomic visual perception: single, low-level judgements about what an image shows rather than multi-step visual reasoning. Its items are not public. This board collects scores reported by model providers and unverified by llm-stats, higher is better. Source: llm-stats.com/benchmarks/perceptionbench.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Around February 2027. 5 models tracked. Updated 21 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Muse Ember | 80.6 | #52 |
| 2 | Kimi K3.1 Pro | 68.3 | #36 |
| 3 | Qwen 3.8 Max | 63.5 | #63 |
| 4 | Kimi K3 | 58.5 | #61 |
| 5 | Inkling 2 Mini | 57.5 | #187 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-perceptionbench · How It Works · Data refreshed daily, snapshot 2026-11-21.