BALSAM - Translation/Transliteration: leaderboard
Metric: Overall score (0-100, LLM-judged generation and multiple choice). Source: benchmarks.ksaa.gov.sa. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemma 4 31B (IT) | 79.65 |
| 2 | GLM-5 (Thinking) | 70.18 |
| 3 | DeepSeek V4 Pro (Reasoning) | 68.53 |
| 4 | GPT-5.2 | 68 |
| 5 | Grok 4.1 Fast (Reasoning) | 67 |
| 6 | DeepSeek V3.2 | 65.16 |
| 7 | Command A 03 2025 | 64.95 |
| 8 | Kimi K2 Instruct (0905) | 63.73 |
| 9 | GLM-4.7 (Reasoning) | 63.61 |
| 10 | Command-R+ | 63.49 |
| 11 | Fanar-C-2-27B | 63.16 |
| 12 | Qwen 3 235B A22B | 61.59 |
| 13 | Claude Opus 4.5 | 60.91 |
| 14 | MiniMax M3 (Reasoning) | 60.78 |
| 15 | AceGPT-v2-8B-Chat | 59.33 |
Interactive version: theaggregate.ai/benchmark?slug=balsam-translation-transliteration · How It Works · Data refreshed daily, snapshot 2026-09-19.