L3Cube-IndicQuest v2 (English): leaderboard
L3Cube-IndicQuest v2 (L3Cube, 2026) tests India-specific factual knowledge with 3,471 curriculum-grounded questions in nine domains, answered closed-book with one to five words and judged by Gemma 3 12B against the gold answer. This board is size-weighted accuracy on the English questions in percent; higher is better.
Metric: Weighted accuracy (%; all 3,471 English questions, domains weighted by size; closed-book short answers to curriculum-grounded India-specific factual questions in English, judged correct or incorrect by Gemma 3 12B against the gold answer (paraphrases and transliterations accepted)). Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 86.9 | #453 |
| 2 | Gemma 4 31B | 62.2 | #394 |
| 3 | GPT-5.4 Mini | 60.1 | #389 |
| 4 | Sarvam 30B | 48.7 | #1744 |
| 5 | Gemma 2 9B | 41.1 | #2802 |
| 6 | Llama 3.1 8B | 35.4 | #4861 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=l3cube-indicquest-v2-english · How It Works · Data refreshed daily, snapshot 2026-11-13.