AURORA Multilingual Terminal-Bench (Hindi)
Agents · 2026-09-30
The 33 Hindi-language tasks of LILT's 324-task AURORA Multilingual Terminal-Bench, with instructions authored by native speakers and run in Harbor with the Terminus-2 agent. The score is the trial-level pass rate pooled over a model's attempts on these tasks.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 75.2 |
| Claude Opus 5 | 75.2 |
| Gemini 3.8 Flash | 62.6 |
| GPT-5.6 Sol | 59.4 |
| GPT-6 Sol | 59.4 |
| Gemini 3.5 Flash | 52.5 |
| Muse Spark 1.3 | 51.5 |
| GPT-5.5 | 49.9 |
| Opus 4.8 | 37.9 |
| Mistral Medium 3.5 | 7.1 |