Open Portuguese LLM - FaQuAD NLI: leaderboard
Deciding whether a sentence answers a question, built from the FaQuAD Portuguese reading-comprehension dataset. From the final table of the archived Open Portuguese LLM Leaderboard (15-shot, lm-evaluation-harness), covering about 1,210 open checkpoints and 42 API models; macro F1 (%), higher is better. Frozen since the last evaluation on 2025-09-01.
Metric: Macro F1 (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 1253 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 (Minimal) | 90.49 | #105 (GPT-5) |
| 2 | Llama 3.3 70B Instruct | 87.88 | #541 |
| 3 | Gemini 2.5 Pro (Preview 03-25) | 87.39 | #126 |
| 4 | Qwen 2.5 32B | 86.86 | #544 |
| 5 | glm-4-9B (0414) | 86.86 | #770 |
| 6 | GPT-4o (2024-08-06) | 86.54 | #343 |
| 7 | Llama 3.2 90B Vision Instruct | 86.32 | #621 |
| 8 | Llama 3.1 70B Instruct | 86.2 | #567 |
| 9 | Mistral-Small-Instruct-2409-abliterated | 86.08 | #809 |
| 10 | Qwen 3 14B | 85.82 | #542 |
| 11 | Gemini 2.5 Flash | 85.79 | #255 |
| 12 | Llama 4 Scout Instruct | 85.67 | #547 |
| 13 | Qwen2.5-7B-Instruct-abliterated-v2 | 85.45 | #660 |
| 14 | Josiefied-Qwen2.5-7B-Instruct-abliterated-v2 | 85.38 | #748 |
| 15 | Kimi K2 | 85.31 | #252 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-portuguese-llm-faquad-nli · How It Works · Data refreshed daily, snapshot 2026-11-05.