⚡ Each donation = another big MoE quantized
I host 30+ 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.
Darwin-36B-Opus — APEX GGUF
APEX (Adaptive Precision for EXpert Models) quantizations of FINAL-Bench/Darwin-36B-Opus.
Brought to you by the LocalAI team | APEX Project | Technical Report
Available Files
| File | Profile | Size | Best For |
|---|---|---|---|
| Darwin-36B-Opus-APEX-I-Balanced.gguf | I-Balanced | 24 GB | Best overall quality/size ratio |
| Darwin-36B-Opus-APEX-Balanced.gguf | Balanced | 24 GB | General purpose |
| Darwin-36B-Opus-APEX-I-Quality.gguf | I-Quality | 22 GB | Highest quality with imatrix |
| Darwin-36B-Opus-APEX-Quality.gguf | Quality | 22 GB | Highest quality standard |
| Darwin-36B-Opus-APEX-I-Compact.gguf | I-Compact | 16 GB | Consumer GPUs, best quality/size |
| Darwin-36B-Opus-APEX-Compact.gguf | Compact | 16 GB | Consumer GPUs |
| Darwin-36B-Opus-APEX-I-Mini.gguf | I-Mini | 13 GB | Smallest "safe" tier |
| Darwin-36B-Opus-APEX-I-Nano.gguf | I-Nano | 11 GB | Experimental — IQ2_XXS mid-layer experts |
| Darwin-36B-Opus-F16.gguf | F16 reference | 65 GB | Full-precision reference |
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).
The key insight: in MoE models, expert FFN tensors make up the bulk of model weight but only ~8/256 experts activate per token. APEX compresses middle-layer experts more aggressively while preserving edge layers (first/last 5) and keeping attention, SSM/Mamba, and shared expert tensors at higher precision.
See the APEX project for full details, technical report, and scripts.
Nano (experimental tier)
The APEX Nano tier pushes mid-layer routed experts to IQ2_XXS (2.06 bpw), near-edge to IQ2_S, edges to Q3_K, with shared experts kept at Q5_K. About 20% smaller than Mini with modest quality cost — viable only on MoE thanks to sparse per-token expert activation. Requires imatrix.
Benchmarks pending. Feedback welcome.
Architecture
- Base: Qwen 3.5 MoE (Qwen3_5MoeForCausalLM) — evolutionary-merge reasoning fine-tune
- Layers: 40
- Experts: 256 routed (8 active per token)
- Total Parameters: ~36B
- Active Parameters: ~3B per token
- Hidden size: 2048
- Attention: Hybrid (full attention every 4th layer, linear/Mamba otherwise)
- APEX Config: 5+5 symmetric edge gradient across 40 layers
- Calibration: v1.3 diverse dataset (chat, code, reasoning, multilingual, tool-calling, Wikipedia)
Run with LocalAI
local-ai run mudler/Darwin-36B-Opus-APEX-GGUF@Darwin-36B-Opus-APEX-I-Balanced.gguf
Credits
- Base / evolutionary merge: FINAL-Bench
- APEX quantization: LocalAI team
- Built on llama.cpp