BALSAM - Question Answering: 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) | 84.39 |
| 2 | GPT-5.2 | 73.98 |
| 3 | GLM-4.7 (Reasoning) | 73.95 |
| 4 | GLM-5 (Thinking) | 73.39 |
| 5 | Command A 03 2025 | 72.2 |
| 6 | DeepSeek V3.2 | 71.65 |
| 7 | Fanar-C-2-27B | 70.35 |
| 8 | AceGPT-v2-8B-Chat | 69.83 |
| 9 | Kimi K2 Instruct (0905) | 68.97 |
| 10 | Llama 3.3 70B Instruct | 67.31 |
| 11 | MiniMax M3 (Reasoning) | 67.24 |
| 12 | Llama 4 Maverick Instruct | 65.44 |
| 13 | Jais-2-70B-Chat | 65.17 |
| 14 | Qwen 3 235B A22B | 60.48 |
| 15 | Grok 4.1 Fast (Reasoning) | 59.91 |
Interactive version: theaggregate.ai/benchmark?slug=balsam-question-answering · How It Works · Data refreshed daily, snapshot 2026-09-19.