Falcon3-7B-Base: benchmark results
TII's Falcon3-7B-Base, a raw pretrained model in the 1B-10B Falcon3 family, trained on 14 trillion tokens with a 32K context and support for English, French, Spanish and Portuguese. Provider: TII. Released 2024-12-17. Access: Open.
Aggregate Rating 1377 ± 22, rank #1310 of 1656 rated models, from 231 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MuSR | 47.02 | Score | 91.5 |
| Open LLM Leaderboard - GPQA | 34.65 | Score | 86.7 |
| EuroEval Spanish NLU - MLQA ES | 65.37 | Character F1 (%) | 84.2 |
| Open Portuguese LLM - OAB Exams | 51.25 | Accuracy (%) | 79.5 |
| Open Portuguese LLM - BLUEX | 60.08 | Accuracy (%) | 77.4 |
| EuroEval Portuguese NLU - MultiWikiQA PT | 73.64 | Character F1 (%) | 76.7 |
| Open Portuguese LLM - ENEM | 70.47 | Accuracy (%) | 76.6 |
| Swallow - Pre-trained English - GSM8K | 76.6 | Exact Match (%) | 75.5 |
| EuroEval Portuguese NLU - SST-2 PT | 81.44 | Matthews correlation coefficient (%) | 74.2 |
| Open LLM Leaderboard - MMLU-Pro | 39.1 | Score | 71 |
| Open LLM Leaderboard - MATH Level 5 | 19.41 | Score | 70 |
| Swallow - Pre-trained English - MATH | 43.8 | Exact Match (%) | 68.4 |
Interactive version: theaggregate.ai/model?slug=falcon3-7b-base · How It Works · Data refreshed daily, snapshot 2026-10-20.