Terminal-Bench-LILT - Arabic: leaderboard
Part of Terminal-Bench-LILT (Kim et al., LILT, 2026): 30 terminal coding tasks written in Arabic by native-speaker programmers around problems specific to non-English software development, such as internationalization, encodings, text normalization and local conventions. Percentage of five trials per task that pass the task tests in the terminus-2 harness; higher is better.
Metric: Pass rate (%; share of 150 trials, five per task, on the 30 Arabic-language terminal coding tasks written by native-speaker programmers, each verified by its task tests; terminus-2 harness, provider-default reasoning; failed trials caused by infrastructure were re-run). Source: arxiv.org. Saturation forecast: Around February 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 69.3 | #49 |
| 2 | Gemini 3.5 Flash | 62.7 | #101 |
| 3 | Claude Opus 4.8 | 59.3 | #56 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=terminal-bench-lilt-arabic · How It Works · Data refreshed daily, snapshot 2026-11-13.