INDICA: leaderboard
INDICA (Common to Whom? Regional Cultural Commonsense and LLM Bias in India, Madhusudan et al., ACL 2026) asks 1,630 region-anchored short-answer questions about everyday cultural practice in North, South, East, West or Central India, with gold answers agreed by annotators from that region. A Gemini 3 Flash judge marks each answer fully correct, partially correct or incorrect over 30 runs at temperature 1.0, and the score is the fully correct rate plus half the partially correct rate, in percent; higher is better.
Metric: Region-Anchored Short Answer Accuracy (%; fully correct plus half credit for partially correct, Gemini 3 Flash judge, 30 runs). Source: arxiv.org. Saturation forecast: Around 2030. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 Fast | 52.6 | #438 |
| 2 | GPT-5.2 | 52.4 | #172 |
| 3 | Claude Sonnet 4.5 | 51.7 | #244 |
| 4 | Qwen 3 VL 235B A22B Instruct | 51.6 | #521 |
| 5 | Gemini 3 Flash (Preview) | 51.1 | #141 |
| 6 | Mistral Large 3 | 50.4 | #819 |
| 7 | Llama 3.3 70B Instruct | 50.3 | #1353 |
| 8 | DeepSeek V3.2 | 49.5 | #403 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=indica · How It Works · Data refreshed daily, snapshot 2026-11-12.