Open Portuguese LLM - ASSIN2 STS: leaderboard
Semantic textual similarity task from the Open Portuguese LLM Leaderboard (ASSIN2), scoring Pearson correlation (x100) between model and human similarity judgments on Brazilian Portuguese sentence pairs.
Metric: Pearson Correlation (×100). Source: huggingface.co. Saturation forecast: Not forecast. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 87.15 | #245 |
| 2 | DeepSeek V2 Chat | 85.33 | #563 |
| 3 | Gemini 2.0 Flash Lite (001) | 84.92 | #544 |
| 4 | Gemini 2.0 Flash (001) | 84.4 | #360 |
| 5 | Gemini 2.5 Flash Lite | 84 | #422 |
| 6 | Gemini 1.5 Flash (001) | 83.88 | #577 |
| 7 | Gemini 1.5 Flash (002) | 83.8 | #583 |
| 8 | Gemini 2.5 Pro (Preview 03-25) | 83.78 | #123 |
| 9 | Qwen 2.5 72B Instruct | 82.31 | #454 |
| 10 | GPT-5 Mini (2025-08-07) | 81.52 | #172 |
| 11 | Gemma 3 27B (IT) | 81.48 | #523 |
| 12 | DeepSeek V3 (0324) | 81.46 | #340 |
| 13 | Claude 3.7 Sonnet (20250219) | 80.88 | #201 |
| 14 | DeepSeek V3.1 | 80.82 | #264 |
| 15 | GPT-4o (2024-08-06) | 80.79 | #336 |
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-assin2-sts · How It Works · Data refreshed daily, snapshot 2026-10-20.