TestsBenchmarks · Multimodal
LVBench
Extreme long video understanding (no tools).
Measured by independent testersIndependent Reported by the makerVendor-reported
| Thinking levelSetting | |||||||
|---|---|---|---|---|---|---|---|
| 1 | Gemini 4 Argon | Google (Gemini / DeepMind) | 91.7% | Maxhighest thinking settings | Maker's own figureVendor-reportedGoogle ↗ | — | 30 Sep 2026 |
| 2 | Gemini 3.8 Flash | Google (Gemini / DeepMind) | 87.1% | Defaultdefault settings (API default thinking_level=medium) | Maker's own figureVendor-reportedGoogle ↗ | — | 2 Sep 2026 |
| 3 | Gemini 3.7 Flash | Google (Gemini / DeepMind) | 85.4% | Defaultdefault settings (API default thinking_level=medium) | Maker's own figureVendor-reportedGoogle ↗ | — | 13 Aug 2026 |
| 4 | Gemini 3.6 Flash | Google (Gemini / DeepMind) | 84.2% | Defaultdefault settings (API default thinking_level=medium) | Maker's own figureVendor-reportedGoogle ↗ | — | 13 Aug 2026 |