BALSAM - Entailment: leaderboard
BALSAM Arabic benchmark (SDAIA/KSAA, dataset version 2), Entailment category: deciding whether a sentence can be logically inferred from another, testing the semantic relation between text and hypothesis. The site's overall score blends LLM-judged generated answers with multiple-choice answers across the category's tasks; 0-100, higher is better.
Metric: Overall score (0-100, LLM-judged generation and multiple choice). Source: benchmarks.ksaa.gov.sa. Saturation forecast: Not forecast. 39 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 87.75 | #115 |
| 2 | Llama 4 Maverick Instruct | 86.92 | #467 |
| 3 | Nova Pro (v1) | 86.25 | #508 |
| 4 | GLM-5 (Thinking) | 83 | #149 (GLM-5) |
| 5 | Yehia-7B-preview | 79.67 | #659 |
| 6 | GPT-OSS-120B (Reasoning) | 79.5 | #346 (GPT-OSS-120B) |
| 7 | AceGPT-v2-8B-Chat | 78.83 | #746 |
| 8 | GLM-4.7 (Reasoning) | 78.42 | #202 (GLM-4.7) |
| 9 | Llama 3.3 70B Instruct | 77.75 | #541 |
| 10 | Qwen 3 235B A22B | 77.08 | #323 |
| 11 | ALLaM-7B-Instruct-preview | 75.75 | #724 |
| 12 | Claude Opus 4.5 | 75.58 | #87 |
| 13 | Jais-2-70B-Chat | 73.17 | #430 |
| 14 | Fanar-C-2-27B | 71.17 | #637 |
| 15 | Command A 03 2025 | 70 | #530 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=balsam-entailment · How It Works · Data refreshed daily, snapshot 2026-11-05.