Kaggle IndicGenBench - XorQA-In: leaderboard
XorQA-In from Google DeepMind's IndicGenBench on Kaggle: answer a question asked in one of 28 Indic languages using an English passage, with bilingual answer annotations. Scored with SQuAD-style token F1 as in the IndicGenBench paper, 0-100, higher is better.
Metric: Token F1. Source: www.kaggle.com. Saturation forecast: Around February 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 76.3 | #71 |
| 2 | Claude Sonnet 4.5 | 75.95 | #148 |
| 3 | Gemini 2.5 Flash | 70.86 | #255 |
| 4 | Grok 4 | 68.34 | #189 |
| 5 | GPT-5 | 67.63 | #105 |
| 6 | O3 | 67.25 | #132 |
| 7 | Claude Opus 4.1 | 66.99 | #146 |
| 8 | Qwen 3 235B A22B 2507 Instruct | 65.15 | #315 |
| 9 | GPT-4.1 | 61.04 | #258 |
| 10 | O4 Mini | 58.12 | #185 |
| 11 | Sarvam M | 54.64 | #643 |
| 12 | GPT-OSS-120B | 53.25 | #346 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=kaggle-indicgenbench-xorqa-in · How It Works · Data refreshed daily, snapshot 2026-11-05.