---
library_name: transformers
license: cc-by-4.0
base_model: Qwen/Qwen3.5-0.8B-Base
language:
- af # Afrikaans
- am # Amharic
- ar # Arabic
- en # English
- fr # French
- ha # Hausa
- ig # Igbo
- mg # Malagasy (Plateau)
- ny # Nyanja
- om # Oromo
- pt # Portuguese
- rw # Kinyarwanda
- sn # Shona
- so # Somali
- st # Southern Sotho
- sw # Swahili
- ti # Tigrinya
- tn # Tswana
- xh # Xhosa
- yo # Yoruba
- zu # Zulu
- rn # Rundi
- lg # Ganda
- ts # Tsonga
- ln # Lingala
- ee # Ewe
- wo # Wolof
- sg # Sango
- ak # Akan
- tw # Twi
- kbp # Kabiye
- bm # Bambara
- nso # Northern Sotho
- fon # Fon
- ss # Swati
- tzm # Central Atlas Tamazight
- kab # Kabyle
- kea # Kabuverdianu
- nqo # N'Ko
- mos # Mossi
- kmb # Kimbundu
- knc # Kanuri
- dyu # Dyula
- taq # Tamasheq
- dik # Southwestern Dinka
- luo # Luo
- ff # Nigerian Fulfulde
- bem # Bemba
- ki # Kikuyu
- kam # Kamba
- kg # Kikongo
- lua # Luba-Kasai
pipeline_tag: text-generation
tags:
- african-languages
- multilingual
- continued-pretraining
- afrique-llm
- qwen3_5
- image-text-to-text
- llamafactory
---
# AfriqueQwen3.5-0.8B-50Langs
## Model Overview
**AfriqueQwen3.5-0.8B-50Langs** is part of the **AfriqueLLM** suite, a collection of open language models adapted to **50 African languages** through continued pre-training (CPT) on **~35.5B tokens**. This model is based on [Qwen/Qwen3.5-0.8B-Base](https://huggingface.co/Qwen/Qwen3.5-0.8B-Base) and has been specifically adapted for improved performance on African languages while maintaining strong capabilities in high-resource languages.
This compact variant follows the same extended code, math, and 50-language continued-pretraining recipe as AfriqueQwen3.5-4B-50Langs.
### Key Features
- **Type**: Causal Language Model (Base/Pre-trained)
- **Base Model**: Qwen 3.5 0.8B
- **Parameters**: 0.8B
- **Context Length**: 262,144 tokens (native)
- **Training Tokens**: ~35.5B tokens of carefully curated multilingual data
## Supported Languages
AfriqueQwen3.5-0.8B-50Langs has been adapted for the following 50 African languages:
| Language | Code | Family | Script |
|----------|------|--------|--------|
| Afrikaans | afr_Latn | Germanic | Latin |
| Swahili | swh_Latn | Bantu | Latin |
| Moroccan Arabic | ary_Arab | Semitic | Arabic |
| Somali | som_Latn | Cushitic | Latin |
| Amharic | amh_Ethi | Semitic | Ethiopic |
| Egyptian Arabic | arz_Arab | Semitic | Arabic |
| Hausa | hau_Latn | Chadic | Latin |
| Kinyarwanda | kin_Latn | Bantu | Latin |
| Zulu | zul_Latn | Bantu | Latin |
| Igbo | ibo_Latn | Volta-Niger | Latin |
| Plateau Malagasy | plt_Latn | Austronesian | Latin |
| Xhosa | xho_Latn | Bantu | Latin |
| Shona | sna_Latn | Bantu | Latin |
| Yoruba | yor_Latn | Volta-Niger | Latin |
| Nyanja | nya_Latn | Bantu | Latin |
| Southern Sotho | sot_Latn | Bantu | Latin |
| Tigrinya | tir_Ethi | Semitic | Ethiopic |
| Tunisian Arabic | aeb_Arab | Semitic | Arabic |
| Oromo | gaz_Latn | Cushitic | Latin |
| Tswana | tsn_Latn | Bantu | Latin |
| Rundi | run_Latn | Bantu | Latin |
| Ganda | lug_Latn | Bantu | Latin |
| Tsonga | tso_Latn | Bantu | Latin |
| Lingala | lin_Latn | Bantu | Latin |
| Ewe | ewe_Latn | Kwa | Latin |
| Wolof | wol_Latn | Senegambian | Latin |
| Sango | sag_Latn | Creole | Latin |
| Akan/Twi | aka_Latn / twi_Latn | Kwa | Latin |
| Kabiye | kbp_Latn | Gur | Latin |
| Bambara | bam_Latn | Mande | Latin |
| Northern Sotho | nso_Latn | Bantu | Latin |
| Fon | fon_Latn | Kwa | Latin |
| Swati | ssw_Latn | Bantu | Latin |
| Central Atlas Tamazight | tzm_Tfng | Berber | Tifinagh |
| Kabyle | kab_Latn | Berber | Latin |
| Kabuverdianu | kea_Latn | Creole | Latin |
| N'Ko | nqo_Nkoo | Mande | N'Ko |
| Mossi | mos_Latn | Gur | Latin |
| Kimbundu | kmb_Latn | Bantu | Latin |
| Kanuri | knc_Arab / knc_Latn | Saharan | Arabic/Latin |
| Dyula | dyu_Latn | Mande | Latin |
| Tamasheq | taq_Latn | Berber | Latin |
| Southwestern Dinka | dik_Latn | Nilotic | Latin |
| Luo | luo_Latn | Nilotic | Latin |
| Nigerian Fulfulde | fuv_Latn | Senegambian | Latin |
| Bemba | bem_Latn | Bantu | Latin |
| Kikuyu | kik_Latn | Bantu | Latin |
| Kamba | kam_Latn | Bantu | Latin |
| Kikongo | kon_Latn | Bantu | Latin |
| Luba-Kasai | lua_Latn | Bantu | Latin |
**High-resource languages used for catastrophic forgetting mitigation:** English, French, Portuguese, Arabic
## Training Data
Our training corpus combines multiple high-quality sources:
- **African Monolingual Data** (~22.8B tokens): FineWeb2, WURA, and MADLAD-400
- **Code** (~1B tokens): CornStack-Python for reasoning capabilities
- **Mathematics** (~1B tokens): FineMath-4+ for mathematical understanding
- **Synthetic Data** (~324M tokens): GPT-4.1 translated domain-specific content across 10 domains
- **Additional Language Expansion** (~1.5B tokens beyond the extended code/math recipe): remaining African language data, upsampled 5x for broader language coverage.
We use **UniMax sampling** to create a balanced distribution, capping high-resource languages at approximately 1B tokens and upsampling lower-resource languages for up to five epochs.
## Quickstart
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "McGill-NLP/AfriqueQwen3.5-0.8B-50Langs"
# Load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype="auto",
device_map="auto"
)
# Prepare the model input
prompt = "Bawo ni o ṣe n ṣe?" # Yoruba: "How are you doing?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
# Generate text
generated_ids = model.generate(
**inputs,
max_new_tokens=100,
)
output = tokenizer.decode(generated_ids[0], skip_special_tokens=True)
print(output)
```
## Deployment
For deployment, you can use `vllm` or `sglang` to create an OpenAI-compatible API endpoint:
**vLLM:**
```shell
vllm serve McGill-NLP/AfriqueQwen3.5-0.8B-50Langs
```
**SGLang:**
```shell
python -m sglang.launch_server --model-path McGill-NLP/AfriqueQwen3.5-0.8B-50Langs
```
## Training Details
### Hyperparameters
- **Learning Rate**: 5e-5 (with warmup and cosine decay)
- **Context Length**: 16,384 tokens
- **Optimizer**: AdamW
- **Precision**: BF16 mixed precision
### Infrastructure
Training was conducted using the LLaMA-Factory framework on up to 64 NVIDIA H100 GPUs with:
- DeepSpeed ZeRO-1/ZeRO-2
- Flash Attention 3
- Sequence packing
- Liger Kernel optimizations
## Evaluation
All AfriqueLLM models are evaluated on multiple multilingual benchmarks. FLORES is reported only in the English-to-target direction (`eng->xxx`):
| Model | AfriMGSM | AfriMMLU | AfriXNLI | Belebele | FLORES (eng->xxx) | INJONG | SIB-200 | Overall | Δ (Δ %) |
|-------|----------|----------|----------|----------|--------|--------|---------|---------|---|
| [Gemma3-4B](https://huggingface.co/google/gemma-3-4b-pt) | 10.24 | 33.89 | 37.76 | 45.79 | 35.36 | 55.52 | 63.59 | 40.31 | |
| [AfriqueGemma-4B](https://huggingface.co/McGill-NLP/AfriqueGemma-4B) | 14.86 | 36.73 | 39.62 | 50.52 | 54.95 | 69.28 | 69.21 | 47.88 | +7.6 (18.8%) |
| [Gemma3-12B](https://huggingface.co/google/gemma-3-12b-pt) | 25.21 | 48.76 | 44.01 | 68.84 | 44.09 | 73.53 | 79.17 | 54.80 | |
| [AfriqueGemma-12B](https://huggingface.co/McGill-NLP/AfriqueGemma-12B) | 32.14 | 49.47 | 44.60 | 68.65 | 65.04 | 76.79 | 75.08 | 58.82 | +4.0 (7.3%) |
| [Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B-Base) | 8.26 | 33.84 | 37.12 | 41.50 | 20.16 | 21.69 | 57.88 | 31.49 | |
| [AfriqueQwen-4B](https://huggingface.co/McGill-NLP/AfriqueQwen-4B) | 33.09 | 43.04 | 44.88 | 63.62 | 59.82 | 65.34 | 74.77 | 54.94 | +23.4 (74.4%) |
| [Qwen3.5-0.8B](https://huggingface.co/Qwen/Qwen3.5-0.8B-Base) | 3.51 | 28.72 | 33.52 | 31.24 | 14.80 | 3.66 | 39.07 | 22.07 | |
| [AfriqueQwen3.5-0.8B-50Langs](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-0.8B-50Langs) | 8.37 | 28.29 | 33.96 | 32.71 | 49.85 | 24.55 | 51.18 | 32.70 | +10.63 (+48.1%) |
| [Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B-Base) | 20.79 | 38.63 | 40.36 | 55.82 | 32.06 | 59.43 | 74.96 | 46.01 | |
| [AfriqueQwen3.5-4B](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B) | 30.47 | 43.66 | 41.05 | 66.01 | 63.55 | 75.46 | 79.66 | 57.12 | +11.1 (24.2%) |
| [AfriqueQwen3.5-4B-ExtendedCM](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B-ExtendedCM) | 34.17 | 45.26 | 41.94 | 66.45 | 63.76 | 75.97 | 80.52 | 58.30 | +1.2 (2.1%) |
| [AfriqueQwen3.5-4B-50Langs](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B-50Langs) | 34.06 | 45.23 | 41.79 | 66.83 | 64.56 | 75.82 | 79.83 | 58.30 | +0.0 (0.0%) |
| [Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B-Base) | 31.61 | 44.00 | 44.14 | 63.87 | 40.80 | 71.06 | 81.12 | 53.80 | |
| [AfriqueQwen3.5-9B-50Langs](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-9B-50Langs) | 40.47 | 49.76 | 45.93 | 72.01 | 65.00 | 78.53 | 82.42 | 62.02 | +8.22 (+15.3%) |
| [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B-Base) | 11.22 | 36.56 | 38.24 | 44.63 | 21.13 | 29.47 | 53.06 | 33.47 | |
| [AfriqueQwen-8B](https://huggingface.co/McGill-NLP/AfriqueQwen-8B) | 39.68 | 46.91 | 45.99 | 68.46 | 62.18 | 73.36 | 77.00 | 59.08 | +25.6 (76.5%) |
| [Qwen3-14B](https://huggingface.co/Qwen/Qwen3-14B-Base) | 16.60 | 39.66 | 43.22 | 50.74 | 23.61 | 41.80 | 66.29 | 40.27 | |
| **[AfriqueQwen-14B](https://huggingface.co/McGill-NLP/AfriqueQwen-14B)** | **45.01** | **52.22** | **49.01** | **74.63** | **63.77** | **77.80** | **82.63** | **63.58** | **+23.3 (57.9%)** |
| [Llama3.1-8B](https://huggingface.co/meta-llama/Llama-3.1-8B) | 8.14 | 32.27 | 37.90 | 40.95 | 26.69 | 41.37 | 59.99 | 35.33 | |
| [AfriqueLlama-8B](https://huggingface.co/McGill-NLP/AfriqueLlama-8B) | 17.51 | 36.57 | 37.39 | 50.51 | 63.60 | 71.17 | 69.14 | 49.41 | +14.1 (39.9%) |
| [Lugha-Llama-8B-wura](https://huggingface.co/Lugha/Meta-Llama-3.1-8B-wura) | 9.46 | 37.00 | 39.24 | 47.86 | 49.90 | 62.30 | 75.81 | 45.94 | |
| [Gemma3-27B](https://huggingface.co/google/gemma-3-27b-pt) | 35.37 | 55.47 | 46.85 | 74.81 | 48.41 | 79.70 | 84.34 | 60.71 | |
### Additional-Language Evaluation
This table averages only evaluated African languages outside the first 20-language CPT set: Ewe, Lingala, Ganda, Twi, and Wolof. Benchmark cells average the available languages for that benchmark; FLORES is English-to-target only (`eng->xxx`).
| Model | AfriMGSM | AfriMMLU | AfriXNLI | Belebele | FLORES (eng->xxx) | INJONG | SIB-200 | Avg | Δ (Δ %) |
|-------|----------|----------|----------|----------|-------------------|--------|---------|-----|--------|
| [Qwen3.5-0.8B](https://huggingface.co/Qwen/Qwen3.5-0.8B-Base) | 2.24 | 26.68 | 33.08 | 28.50 | 15.52 | 5.60 | 35.50 | 21.02 | |
| [AfriqueQwen3.5-0.8B-50Langs](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-0.8B-50Langs) | 3.71 | 25.35 | 33.01 | 28.69 | 27.64 | 14.10 | 42.23 | 24.96 | +3.94 (+18.8%) |
| [Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B-Base) | 8.30 | 32.35 | 34.30 | 36.90 | 22.20 | 33.27 | 58.96 | 32.33 | |
| [AfriqueQwen3.5-4B](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B) | 7.52 | 31.17 | 33.56 | 35.82 | 26.28 | 33.10 | 56.66 | 32.02 | -0.31 (-1.0%) |
| [AfriqueQwen3.5-4B-ExtendedCM](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B-ExtendedCM) | 8.13 | 32.72 | 33.74 | 36.76 | 24.06 | 32.41 | 56.00 | 31.97 | -0.05 (-0.2%) |
| [AfriqueQwen3.5-4B-50Langs](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B-50Langs) | 21.07 | 37.65 | 36.31 | 51.56 | 56.33 | 61.37 | 75.75 | 48.58 | +16.61 (+52.0%) |
| [Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B-Base) | 14.42 | 34.94 | 36.10 | 42.04 | 20.14 | 45.50 | 66.72 | 37.12 | |
| [AfriqueQwen3.5-9B-50Langs](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-9B-50Langs) | 25.30 | 40.06 | 38.27 | 55.58 | 58.93 | 68.60 | 79.52 | 52.32 | +15.20 (+40.9%) |
## Model Variants
- [AfriqueQwen-14B](https://huggingface.co/McGill-NLP/AfriqueQwen-14B) - Qwen-based 14B model (flagship)
- [AfriqueQwen-8B](https://huggingface.co/McGill-NLP/AfriqueQwen-8B) - Qwen-based 8B model
- [AfriqueQwen-4B](https://huggingface.co/McGill-NLP/AfriqueQwen-4B) - Qwen-based 4B model
- [AfriqueQwen3.5-9B-50Langs](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-9B-50Langs) - Qwen 3.5-based 9B model with 50-language coverage (Qwen 3.5 flagship)
- [AfriqueQwen3.5-4B](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B) - Qwen 3.5-based 4B model
- [AfriqueQwen3.5-4B-ExtendedCM](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B-ExtendedCM) - Qwen 3.5-based 4B model with extended continued pre-training
- [AfriqueQwen3.5-4B-50Langs](https://huggingface.co/McGill-NLP/AfriqueQwen3.5-4B-50Langs) - Qwen 3.5-based 4B model with 50-language coverage
- [AfriqueGemma-4B](https://huggingface.co/McGill-NLP/AfriqueGemma-4B) - Gemma-based 4B model
- [AfriqueGemma-12B](https://huggingface.co/McGill-NLP/AfriqueGemma-12B) - Gemma-based 12B model
- [AfriqueLlama-8B](https://huggingface.co/McGill-NLP/AfriqueLlama-8B) - Llama-based 8B model
## Citation
If you find our work helpful, please cite:
```bibtex
@misc{yu2026afriquellmdatamixingmodel,
title={AfriqueLLM: How Data Mixing and Model Architecture Impact Continued Pre-training for African Languages},
author={Hao Yu and Tianyi Xu and Michael A. Hedderich and Wassim Hamidouche and Syed Waqas Zamir and David Ifeoluwa Adelani},
year={2026},
eprint={2601.06395},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2601.06395},
}
```
## License
This model is released under the [CC BY 4.0 License](https://creativecommons.org/licenses/by/4.0/). Please review the license terms before use.
## Acknowledgments
We thank the creators of the base models, datasets and compute resources that made this work possible, including Mila, Compute Canada, Microsoft, the FineWeb team, WURA, MADLAD-400 and etc..