Multimodal Computer Use on AOI 12-task audio
3Pod CountAOI full (Claude 4.6)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| AOI full (Claude 4.6)2026.06 | 3 | 3 | 3 | 3 | 12 | 100 | |
| gpt-realtime-2mode=+ AOI scaffold2026.06 | 2 | 3 | 3 | 3 | 11 | 92 | |
| Grok Voice (no vision)mode=audio only2026.06 | 1 | 0 | 0 | 0 | 1 | 8 | |
| Grok Voice (no vision)mode=+ scaffold2026.06 | 1 | 0 | 0 | 0 | 1 | 8 | |
| Gemini Live (2.5)2026.06 | 0 | 0 | 0 | 0 | 0 | 0 | |
| OpenAI Realtime (gpt-4o)2026.06 | 0 | 0 | 0 | 3 | 3 | 25 | |
| gpt-realtime-2mode=alone2026.06 | 0 | 0 | 1 | 1 | 2 | 17 |