FilBench - Classical NLP: leaderboard
From the FilBench leaderboard (Miranda et al., 2025), which evaluates LLMs on Filipino, Tagalog and Cebuano with a lighteval fork. Classical NLP combines BalitaNLP, CebuaNER, Dengue, FiReCS, SIB-200, TLUnified NER and Universal NER as multiple-choice tasks, weighted by instance count; accuracy 0-100, higher is better. Runs missing one of these tasks are omitted.
Metric: Category Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 37 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.1 70B Instruct | 90.27 | #567 |
| 2 | Llama-SEA-LION-v3-70B-IT | 89.99 | #407 |
| 3 | Qwen 2.5 32B Instruct | 89.32 | #509 |
| 4 | GPT-4o (2024-08-06) | 89.03 | #343 |
| 5 | Sailor2-20B-Chat | 89.03 | #589 |
| 6 | c4ai-command-a-03-2025 | 88.97 | #468 |
| 7 | Gemma 3 27B (IT) | 88.61 | #528 |
| 8 | Qwen 2.5 72B Instruct | 88.6 | #469 |
| 9 | Gemma 3 12B (IT) | 88.56 | #672 |
| 10 | gemma2-9B-cpt-sea-lionv3-instruct | 88.55 | #717 |
| 11 | Gemma 2 27B (IT) | 87.99 | #652 |
| 12 | Llama 4 Scout Instruct | 87.88 | #547 |
| 13 | Gemma 2 9B (IT) | 87.47 | #795 |
| 14 | aya-expanse-32B | 87.47 | #687 |
| 15 | Llama 4 Maverick Instruct FP8 | 87.28 | #424 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=filbench-classical-nlp · How It Works · Data refreshed daily, snapshot 2026-11-05.