Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Get help setting up a custom Dedicated Endpoints.
Talk with our engineer to get a quote for reserved GPU instances with discounts.
README
License: apache-2.0Overview
Averroes-7B-v1 is the original fine-tuned Arabic-English model built on Qwen2.5-7B-Instruct. It was the first iteration trained on the full 1.3M Arabic instruction corpus.
| Detail | Value |
|---|---|
| Base model | Qwen/Qwen2.5-7B-Instruct |
| Training data | 1.3M Arabic instruction pairs (full corpus) |
| Training duration | ~50 hours on M2 Ultra |
| License | Apache 2.0 |
🏆 Current Champion
For better performance and faster inference, use Averroes-v1.5-fused:
→ hayulalab/averroes-v2-instruct-v1.5-fused
Improvements over v1:
- ✅ +30% English HumanEval (86.7% vs 56.7%)
- ✅ +15.2% Arabic HumanEval (46.9% vs 31.7%)
- ✅ 3.4× faster inference on Apple Silicon
- ✅ GGUF quantizations available (Q4_K_M, Q8_0, FP16)
- ✅ Clean model card with full benchmarks
Links
- Champion model: https://huggingface.co/hayulalab/averroes-v2-instruct-v1.5-fused
- Hayula Models: https://github.com/HayulaLab/hayula-models
- UTA Platform: https://github.com/HayulaLab/uta
- Website: https://hayula.xyz
- Organization: https://huggingface.co/hayulalab
Part of Hayula — Open-Source Arabic AI Lab. Built with ❤️.
Model provider
hayulalab
Model tree
Base
Qwen/Qwen2.5-7B-Instruct
Quantized
this model
Modalities
Input
Text
Output
Text
Pricing
Dedicated Endpoints
View detailsSupported Functionality
Model APIs
Dedicated Endpoints
Container
More information