Views
No views yet
| Property | Value |
|---|---|
| Architecture | Qwen3.5-MoE (Mixture of Experts) |
| Parameters | ~35B total |
| Hidden size | 2048 |
| Layers | 40 (linear attention + full attention hybrid) |
| Experts | 256 + 1 shared, 8 active per token |
| Context length | 262,144 tokens |
| Vocabulary | 248,320 |
| RoPE theta | 10,000,000 |
| Base model | osunlp/QUEST-35B-RL |
| File | Size | BPW | Description |
|---|---|---|---|
QUEST-35B-RL-Q6_K.gguf | 27 GB | 6.58 | High quality, recommended |
QUEST-35B-RL-Q4_K_M.gguf | 20 GB | 4.88 | Good quality, smaller footprint |
1# llama.cpp
2./llama-cli -m QUEST-35B-RL-Q6_K.gguf -ngl 999 -c 262144
3
4# llama-server
5./llama-server -m QUEST-35B-RL-Q6_K.gguf -ngl 999 -c 262144 --port 8080convert_hf_to_gguf.py and quantized with llama-quantize.