Benchmark
Indigenous Americas Language Identification Hard (zero-shot)
Scores are percentages. Higher is better.
- Models measured
- 7
- Scores recorded
- 7
- Sources
- 1
- IQ weight
- 0.90
Score distribution
- 60.9%
- Mean
- 55%
- Median
- 13.9%
- Std dev
- 51%
- 10th percentile
- 78%
- 90th percentile
- 41%
- Min
This benchmark has a weight of 0.90 for chat models, measured over 7 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | GPT 4.1table: 1, mode: Hard· Indigenous Americas linguistic competence data docs | 81% |
| 2 | Gemini 2.0 Flashtable: 1, mode: Hard· Indigenous Americas linguistic competence data docs | 78% |
| 3 | DeepSeek V3table: 1, mode: Hard· Indigenous Americas linguistic competence data docs | 68% |
| 4 | Qwen 2.5 14B Instructtable: 1, mode: Hard· Indigenous Americas linguistic competence data docs | 55% |
| 5 | Llama 3.1 8B Instructtable: 1, mode: Hard· Indigenous Americas linguistic competence data docs | 52% |
| 6 | Qwen2.5 7B Instructtable: 1, mode: Hard· Indigenous Americas linguistic competence data docs | 51% |
| 7 | Qwen2.5 3B Instructtable: 1, mode: Hard· Indigenous Americas linguistic competence data docs | 41% |
Data credits: Vasselli, Martínez Peguero, Hudi, Sakajo and Watanabe. Measuring Linguistic Competence of LLMs on Indigenous Languages of the Americas. EACL 2026, ©2026 ACL, CC BY 4.0. Numeric/protocol extract from Tables 1–4; random reference rows omitted. (licence)