Update README.md

README.md

@@ -302,7 +302,7 @@ Users should:
 
 ### Environmental Impact
 
-- **Training**: ~6-8 hours on 2x A100 GPUs
+- **Training**: ~6-8 hours on 1x RTX PRO 6000 GPUs
 - **Carbon Footprint**: Estimated ~5-10 kg CO2eq (depends on energy source)
 - **Inference**: Efficient at 1.3B parameters, suitable for edge deployment
 
@@ -371,7 +371,7 @@ The model uses the **Asterisk** architecture, which combines:
 - ✅ **Full Hybrid**: All 16 layers use ASPP + Attention
 - ✅ **Bilingual**: Serbian + English capabilities
 - ✅ **Reasoning**: Math, code, and general reasoning
-- ✅ **Fast Training**: ~6-8 hours on 2x A100
+- ✅ **Fast Training**: ~6-8 hours on 1x RTX PRO 6000
 - ✅ **Low Memory**: ~3GB inference, ~20GB training per GPU
 
 ## Hardware Requirements