Zero-shot video understanding on VideoZeroBench
27.6Level 3 ScoreGPT-5.4 xHigh
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5.4 xHighHarness=Codex, Reasoning effort=xHigh2026.05 | 27.6 | |
| GPT-5.4 HighHarness=Codex, Reasoning effort=High2026.05 | 27.2 | |
| GPT-5.4 MediumHarness=Codex, Reasoning effort=Medium2026.05 | 24 | |
| GPT-5.4 LowHarness=Codex, Reasoning effort=Low2026.05 | 21 | |
| GPT-5.4 High§Harness=Codex, Direct image read=disabled2026.05 | 19 | |
| Gemini 3 Flash2026.05 | 17.8 | |
| Gemini 3.1 Pro2026.05 | 17.6 | |
| Kimi K-2.5Harness=CC2026.05 | 15.6 | |
| OmniAgentBackbone=GPT-5.4 High2026.05 | 12.2 | |
| Qwen 3.5 Omni Plus2026.05 | 10.4 | |
| Mimo V2 Omni2026.05 | 7.4 | |
| MiniMax-M2.7‡Harness=CC2026.05 | 5.8 | |
| Qwen 3 Omni2026.05 | 5.2 | |
| Agent-OmniBackbone=GPT-5.4 High2026.05 | 5 |