Quick Navigation Index

Independently computed using dedicated cloud compute clusters, funded 100% out-of-pocket.

If you consider these releases to be of high quality and they help you in your day-to-day workflow, consider supporting on Ko-fi. Every contribution directly sustains future cloud compute hours and helps support my university studies.

1. APEX MLX Variants
2. Fidelity Verification
3. Quality Spectrum: APEX MLX NanoPlus vs. MiniPlus
4. Model Files & Technical Specifications
5. Surgical Tensor Quantization Map
6. Apple Silicon Unified Memory & Context
7. Recommended Configuration & Setup
8. Recommended Generation Parameters
9. CRITICAL: Coding Syntax & Repeat Penalty Advisory
10. Native Chat Template
11. 7. Community Compute Fund & Student Support

Qwen3.8-35B-A3B-Distill MLX APEX-NanoPlus Abliterated (Uncensored)

Compact Q4-Class Frontier MoE · Native MLX Mixed Precision · Apple Silicon · 18.82 GB Weights

[!WARNING]

Known Upstream Long-Generation Limitation

The upstream Qwen3.8-35B-A3B-Distill checkpoint can repeat during very long responses, including after the <think> block. This behavior originates upstream and is not specific to the MLX APEX quantization.

[!NOTE]

🚀 LOOKING FOR THE HIGHER-FIDELITY MLX RELEASE?

If you can spend about 1.13 GB more storage/unified-memory weight budget, use the sister MiniPlus release: 🔗 Qwen3.8-35B-A3B-Distill MLX APEX-MiniPlus Abliterated — Q5-class practical quality, 19.95 GB of Safetensors weights and a larger precision budget on always-active components.

[!TIP]

🔓 ABLITERATED / UNCENSORED MLX EDITION

This is the Abliterated / Uncensored MLX APEX-NanoPlus edition of Qwen3.8-35B-A3B-Distill. It is the compact member of the two-model MLX APEX family and is intended for users who deliberately prefer reduced refusal behavior.

[!IMPORTANT]

THE COMPACT SPECIFICATION: Q4-CLASS QUALITY AT 18.82 GB

MLX APEX-NanoPlus is a native MLX mixed-precision release built around a 3-bit G32 base while selectively retaining substantially higher precision on sensitive and always-active paths. The result is 18.8156 GB (17.52 GiB) of Safetensors weights with a Q4-class practical-quality target.

[!TIP]

🏆 THE MLX NANOPLUS SWEET SPOT: Q4-CLASS PRACTICAL QUALITY

Why NanoPlus is not equivalent to a flat 3-bit model:

• LM head: 6-bit G64.
• Shared experts: 5-bit G64.
• Linear-attention in_proj_z: 8-bit G64.
• Norms, MoE routers and recurrent state/control tensors: protected in FP32.
• Full-attention Q/K/V: 4-bit G64, with o_proj at 6-bit G64.

The Hub may identify the checkpoint from its 3-bit G32 base quantization, but that label does not describe the precision of the complete tensor map.

[!TIP]

🏆 FIDELITY RESULT

Release	Weight Size	Quality Class	ΔPPL	%PPL
MLX APEX-NanoPlus Abliterated (CURRENT)	18.8156 GB	Q4-class	+0.826288	+9.365%

PPL is a useful fidelity estimate, not a complete measure of real-world reasoning, coding, tool use or long-context behavior.

[!WARNING]

DO NOT CONFUSE APEX MLX MINIPLUS / NANOPLUS WITH GENERIC APEX OR APEX-I GGUF RELEASES!

The MLX APEX family is a separate native MLX mixed-precision specification. It is not a renamed GGUF APEX-I-Mini, APEX-I-MiniPlus, K-quant or IQ-quant.

• Generic / GGUF APEX releases: use llama.cpp/GGUF quantization types such as Q*_K and IQ*, with their own tensor layouts, kernels and bit accounting.
• MLX APEX NanoPlus / MiniPlus: use MLX affine quantization with per-module bit width and group-size overrides, plus protected FP32 paths. There are no IQ quants here, which is why these releases use APEX, not APEX-I.
• Quality-class labels: Q4-class and Q5-class describe practical fidelity positioning; they do not claim that the checkpoints are uniform flat 4-bit or 5-bit models.

[!TIP]

APPLE SILICON UNIFIED MEMORY

MLX uses Apple Silicon unified memory rather than the discrete GPU-VRAM/system-RAM split used by CUDA + llama.cpp. NanoPlus weights alone occupy about 17.52 GiB; runtime state, activations, macOS and the active context require additional unified memory. Context headroom therefore depends on the Mac and its available memory at inference time.

---

APEX MLX Variants

NanoPlus prioritizes the smaller footprint while retaining high precision on sensitive global, routing and recurrent-control paths. MiniPlus spends the additional budget on higher precision across the always-active shared experts and linear-attention projections.

---

Fidelity Verification

The model-specific fidelity result is summarized at the top of the card:

• ΔPPL: +0.826288
• %PPL: +9.365%

These values are useful as a quantization-fidelity estimate. They should not be interpreted as a direct benchmark score for coding, reasoning, agents or instruction following.

Quality Spectrum: APEX MLX NanoPlus vs. MiniPlus

The quality tiers are practical positioning labels across two different MLX mixed-precision maps, not claims of uniform flat-bit equivalence.

---

Model Files & Technical Specifications

• Base Model: empero-ai/Qwen3.8-35B-A3B-Distill
• Architecture: Qwen3.8 / Qwen3.5-MoE family, 40 layers, 256 micro-experts with hybrid linear attention / recurrent state and periodic full-attention layers
• Runtime: MLX / MLX-LM on Apple Silicon
• Context Length: 262,144 tokens (native 256K); usable context is additionally constrained by available unified memory

---

Surgical Tensor Quantization Map

This is the MLX-native mixed-precision map represented by the published checkpoint:

---

Apple Silicon Unified Memory & Context

Unlike a GGUF CUDA deployment, MLX does not split the model into -ngl GPU layers and system-RAM offload. The CPU and GPU share the same unified-memory pool.

No Mac throughput figure is claimed here without a corresponding measured Apple-Silicon result.

---

Recommended Configuration & Setup

pip install -U mlx-lm

mlx_lm.generate \
  --model IsValorum/Qwen3.8-35B-A3B-Distill-MLX-APEX-NanoPlus-Abliterated \
  --prompt "Hello"

Server:

mlx_lm.server --model IsValorum/Qwen3.8-35B-A3B-Distill-MLX-APEX-NanoPlus-Abliterated

⚙️ Recommended Generation Parameters (empero-ai Official)

Official generation guidance used by empero-ai/Qwen3.8-35B-A3B-Distill:

[!IMPORTANT]

🔍 Model Behavior vs. Quantization Fidelity

Sampling, prompting and upstream model behavior can materially affect the output. The Q4-class label describes this release's practical fidelity target; PPL is only one estimate of quantization fidelity and does not guarantee identical real-world behavior to BF16.

[!IMPORTANT]

CRITICAL ADVISORY FOR CODING WORKFLOWS: PREVENTING SYNTAX & TOKEN SWAPPING

Programming code necessarily repeats brackets, operators and indentation. A repeat penalty above 1.0 can suppress legitimate syntax tokens and cause character substitutions.

Recommended coding profile:

• repeat_penalty: 1.0
• presence_penalty: 0.0
• frequency_penalty: 0.0
• temperature: 0.60 (or lower for stricter deterministic code)
• top_p: 0.95
• top_k: 20

[!TIP]

NATIVE CHAT TEMPLATE

chat_template.jinja and the tokenizer configuration are included in this repository. Use the repository's native template with MLX-LM so system, user, assistant, reasoning and tool-call formatting follow the model's expected control-token structure.

7. Community Compute Fund & Student Support

All IsValorum quantizations will always remain completely free and open to the public without paywalls.

However, cloud GPU compute is expensive and funded entirely out-of-pocket. If you find these handcrafted builds to be of high quality and they benefit your day-to-day workflow, you can support future Day-0 releases on Ko-fi. Every contribution directly funds the cloud compute hours required for new model runs and helps support my university education:

(When supporting on Ko-fi, feel free to leave a note with your Hugging Face handle and the specific model you would like prioritized).