Qwen3.5-122B-A10B REAP30 APEX GGUF
Physical REAP30-pruned APEX GGUF builds of Qwen/Qwen3.5-122B-A10B.
Available quants
| File | Description |
|---|---|
Qwen3.5-122B-A10B-REAP30-APEX-Mini.gguf | Production APEX Mini quant with imatrix |
Qwen3.5-122B-A10B-REAP30-APEX-I-Compact.gguf | Higher-quality APEX I-Compact quant with imatrix |
REAP pruning
- Method: physical REAP expert pruning
- Compression ratio: 30%
- Original experts per MoE layer: 256
- Retained experts per MoE layer: 180
- Layers: 48
- Experts per token: 8
- Observation: 32 batches
- Distance: cosine
- Seed: 42
- Calibration source: custom
calibration-v2
APEX Mini quantization
- Tensor config:
configs/qwen35_122b_mini.txt - Base fallback type:
Q3_K_M - Imatrix context:
-c 4096 - Imatrix chunks:
--chunks 128 - Imatrix calibration: 64MB shuffled
calibration-v2 - Imatrix model source: REAP30 BF16 GGUF
- Text-only GGUF: MTP disabled before conversion
The Mini profile uses Q3_K edge experts and IQ2_S middle routed experts, so imatrix is required.
Example: llama.cpp
llama-cli \
-m Qwen3.5-122B-A10B-REAP30-APEX-Mini.gguf \
-p "<|im_start|>user
Привіт. Напиши один короткий параграф українською. /no_think
<|im_end|>
<|im_start|>assistant
" \
-n 128 \
-c 4096 \
-ngl 40 \
--temp 0.6 \
--top-p 0.95
Notes
Experimental physical expert-pruned build for testing and comparison with REAP20/REAP40 variants.
## APEX I-Compact quantization
- Tensor config: `configs/qwen35_122b_compact.txt`
- Base fallback type: `Q4_K_M`
- Imatrix context: `-c 4096`
- Imatrix chunks: `--chunks 128`
- Imatrix calibration: 64MB shuffled `calibration-v2`
- Imatrix model source: REAP30 BF16 GGUF