empero-ai
Qwen3.8 35B A3B Distill
Qwen3.8 35B A3B Distill is a 35.5B mixture-of-experts qwen35moe model from empero-ai, with about 3.0B parameters active per token. Configured context length is 262,144 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters35.5B / 3.0B active
Context256K
Attentionhybrid
Licenseapache-2.0
40 layers · 2 KV heads · head dim 256 · 30 linear-attention layers · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 66.2 GB | Hugging Face file bytes |
| Q8_0 | 35.2 GB | Hugging Face file bytes |
| Q6_K | 27.2 GB | Hugging Face file bytes |
| Q5_K_M | 23.6 GB | Hugging Face file bytes |
| Q4_K_M | 20.2 GB | Hugging Face file bytes |
| IQ4_XS | 18.3 GB | Hugging Face file bytes |
| Q3_K_M | 16.5 GB | Hugging Face file bytes |
| IQ3_M | 15.2 GB | Hugging Face file bytes |
| Q2_K | 12.9 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Expert offload | Q4_K_M | 78–106 |
| GeForce RTX 3060 12GB | Expert offload | Q4_K_M | 97–133 |
| GeForce RTX 4060 Ti 16GB | Tight fit | Q2_K | 118–162 |
| GeForce RTX 4090 | Tight fit | Q4_K_M | 277–378 |
| GeForce RTX 5090 | Tight fit | Q6_K | 374–510 |
| Apple M4 Pro | Tight fit | Q3_K_M | 90–123 |
| Apple M5 | Won't fit | Q2_K | — |
| CPU only · DDR5-5600 | Runs on CPU | Q5_K_M | 21–29 |
Smallest devices that fit it fully
- GeForce RTX 5080 — Tight fit, Q2_K, 395–538 tok/s
- GeForce RTX 5070 Ti — Tight fit, Q2_K, 369–503 tok/s
- GeForce RTX 5060 Ti 16GB — Tight fit, Q2_K, 184–251 tok/s
- GeForce RTX 4080 Super — Tight fit, Q2_K, 303–413 tok/s
- GeForce RTX 4080 — Tight fit, Q2_K, 295–402 tok/s
- GeForce RTX 4070 Ti Super — Tight fit, Q2_K, 276–377 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (Q2_K).
llama.cpp
llama-server -hf empero-ai/Qwen3.8-35B-A3B-Distill-GGUF:Q2_K -c 8192 -ngl 40
Ollama
ollama run qwen3.8
LM Studio
lms get empero-ai/Qwen3.8-35B-A3B-Distill-GGUF # Then load quant Q2_K