⚡ Each donation = another big MoE quantized
I host 25+ free APEX MoE quantizations as independent research. My only local hardware is an NVIDIA DGX Spark (122 GB unified memory), enough for ~30-50B-class MoEs, but bigger ones (200B+) require rented compute on H100/H200/Blackwell, typically $20-100 per quant.
If APEX quants are useful to you, your support directly funds those bigger runs.
🎉 Patreon (Monthly) | ☕ Buy Me a Coffee | ⭐ GitHub Sponsors
💚 Big thanks to Hugging Face for generously donating additional storage, much appreciated.
Qwopus-MoE-35B-A3B APEX GGUF
APEX (Adaptive Precision for EXpert Models) quantizations of Qwopus-MoE-35B-A3B.
Brought to you by the LocalAI team | APEX Project | Technical Report
Available Files
| File | Profile | Size | Best For |
|---|---|---|---|
| Qwopus-MoE-35B-A3B-APEX-I-Balanced.gguf | I-Balanced | TBD | Best overall quality/size ratio (with imatrix) |
| Qwopus-MoE-35B-A3B-APEX-I-Quality.gguf | I-Quality | TBD | Best quality/compression ratio (with imatrix) |
| Qwopus-MoE-35B-A3B-APEX-Quality.gguf | Quality | TBD | Best quality/compression ratio |
| Qwopus-MoE-35B-A3B-APEX-Balanced.gguf | Balanced | TBD | Best absolute quality |
| Qwopus-MoE-35B-A3B-APEX-I-Compact.gguf | I-Compact | TBD | Consumer GPUs (with imatrix) |
| Qwopus-MoE-35B-A3B-APEX-Compact.gguf | Compact | TBD | Consumer GPUs |
| Qwopus-MoE-35B-A3B-APEX-I-Mini.gguf | I-Mini | TBD | Smallest viable |
What is APEX?
APEX is a quantization strategy for Mixture-of-Experts (MoE) models. It classifies tensors by role (routed expert, shared expert, attention) and applies a layer-wise precision gradient -- edge layers get higher precision, middle layers get more aggressive compression. I-variants use diverse imatrix calibration (chat, code, reasoning, tool-calling, agentic traces, Wikipedia).
See the APEX project for full details, technical report, and scripts.
Architecture
- Model: Qwopus-MoE-35B-A3B (qwen3_5_moe)
- Layers: 40 (hybrid: linear attention + full attention every 4th layer)
- Experts: 256 routed (8 active per token)
- Total Parameters: ~35B
- Active Parameters: ~3B per token
- Origin: Claude Opus 4.6 QLoRA distill of Qwen3.5-35B-A3B
- APEX Config: 5+5 symmetric edge gradient across 40 layers
- Calibration: v1.3 diverse dataset (chat, code, reasoning, multilingual, tool-calling, Wikipedia)
- Source: samuelcardillo/Qwopus-MoE-35B-A3B
Run with LocalAI
local-ai run mudler/Qwopus-MoE-35B-A3B-APEX-GGUF@Qwopus-MoE-35B-A3B-APEX-I-Balanced.gguf
Credits
APEX is brought to you by the LocalAI team. Developed through human-driven, AI-assisted research. Built on llama.cpp.