Pointerbench — Pro
Multimodal · 2026-07-02
Pointerbench-Pro is a 500-example synthetic GUI-grounding benchmark for icon, text, and mixed targets across 100 professional applications. A predicted absolute-pixel click is correct when it lands inside the target bounding box.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 80.8 |
| Opus 4.8 | 80.2 |
| Sonnet 4.6 | 77.4 |
| GPT-5.5 | 75.8 |
| Opus 4.7 | 75.4 |
| GPT-5.6 Sol | 74.0 |
| GPT-5.6 Terra Pro | 74.0 |
| GPT-5.6 Sol Pro | 72.0 |
| GPT-5.6 Terra | 70.0 |
| GPT-5.6 Luna Pro | 62.0 |
| GPT-5.6 Luna | 60.0 |
| GPT-5.4 | 50.4 |
| GPT-5 | 20.2 |
| GPT-5 Mini | 5.8 |
| MiniMax M3 | 3.6 |
| Qwen3.6 Plus (2026-04-02) | 3.6 |
| Qwen3 VL 235B A22B Thinking | 3.2 |
| Kimi K2.5 | 3.0 |
| Qwen3.7-Plus | 3.0 |
| Kimi K2.6 | 2.8 |
| Qwen3.5-9B | 2.6 |
| Qwen3.6-Flash-2026-04-16 | 2.6 |
| GPT-5 Nano | 1.6 |
| Kimi K2.7 Code | 0.8 |
| Gemini 3.5 Flash | 0.6 |
| Gemini 3.1 Pro Preview | 0.4 |
| Step 3.7 Flash | 0.2 |
| Grok 4.3 | 0.0 |
| Grok Build 0.1 | 0.0 |
| MiMo-V2.5 | 0.0 |