Qwen3.5-35B-A3B IT UD IQ3_XXS.GGUF
It's the ideal choice for those who value broad reasoning and encyclopedic accuracy over pure instruction-following.
This IQ3_XXS version, calibrated with the fineweb-edu dataset. i compared various Qwen3.5-35B-A3B-UD quantifications on huggingface. according to some of my observations; outperforms the others in complex medical reasoning and diagnostic accuracy (even though fineweb-edu is not a medical specific dataset).
If you're going to use it in a language other than English, I recommend these sampling settings:
| Parameter | Value |
|---|---|
| Temperature | 0.3 - 0.4 |
| Top K | 40 |
| Repeat Penalty | 1.14 (Coding: 1.02) |
| Top P | 0.95 |
| Min P | 0.01 or 0 |
quantized on NVIDIA Q RTX 8000 with 258GB system RAM
- Downloads last month
- 240
Hardware compatibility
Log In to add your hardware
3-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support