CoolFace
Modelpublic

Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DFlash2

sourceHugging Facemitupdated 12d agoView on Hugging Face
2likes6.7kdownloads
Model Card

<p align="center"> <img src="https://cdn-uploads.huggingface.co/production/uploads/67c2e844e0921a5410eec10a/Y5M42dCag2f7Fc6fDtV0Z.jpeg" alt="Solstice-AI Banner" width="100%"> </p>

<h1 align="center">GLM-5.3-Flash-UNCENSORED (NVFP4 W4A16)</h1>

<h3 align="center">Official Solstice-AI W4A16 NVFP4 Release &bull; Native Multimodal Vision + Video &bull; 1M Context Window (1,048,576 Tokens) &bull; Bundled DFlash 2 Speculative Drafter</h3>

<p align="center"> <b>Original Architecture by <a href="https://huggingface.co/zai-org">Zhipu AI / ZAI</a> &bull; Uncensored Weights by <a href="https://huggingface.co/dealignai">dealignai</a> &bull; NVFP4 W4A16 Packaging & Curation by <a href="https://huggingface.co/Solstice-AI">Solstice-AI</a></b> </p>

<p align="center"> <img src="https://img.shields.io/badge/org-Solstice--AI-blueviolet" alt="Solstice-AI"> <img src="https://img.shields.io/badge/license-MIT-blue" alt="License"> <img src="https://img.shields.io/badge/format-NVFP4%20(W4A16)-orange" alt="Format"> <img src="https://img.shields.io/badge/precision-4--bit%20weights%20%7C%2016--bit%20activations-yellow" alt="Precision"> <img src="https://img.shields.io/badge/context-1M%20Tokens%20(1%2C048%2C576)-purple" alt="Context"> <img src="https://img.shields.io/badge/hardware-Blackwell%20%7C%20Hopper%20%7C%20Ada%20%7C%20Ampere-brightgreen" alt="Hardware"> </p>

Model Summary

`Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4` is the official W4A16 NVFP4 mixed-precision release of the 320B foundation model, GLM-5.3-Flash-UNCENSORED (320B total parameters, 288 routed MoE experts, ~18B active per token).

Key Architectural Highlights:

  • —W4A16 Mixed-Precision: Routed MoE experts quantized to NVFP4 (4-bit float, e2m1) while all attention layers, shared experts, and input activations remain strictly in full 16-bit (BF16/FP16). Zero activation clipping degradation!
  • —Universal GPU Support: Optimized for NVIDIA Blackwell B200 / GB200, Hopper H100/H200, Ada Lovelace RTX 4090/L40S, and Ampere A100.
  • —Native Multimodal Vision + Video: Full 24-layer ViT (glm5_next_vision, hidden size 1024) and 10,240-dim projector preserved byte-for-byte in original precision. Handles high-resolution images and temporal video sequences.
  • —Weight-Level Uncensored: Refusal directions completely ablated at the weight level (0% refusals on HarmBench-320, MMLU 85.28% preserved).
  • —Native 1M Context Window: 1,048,576 tokens native context.
  • —Bundled DFlash 2 Speculative Drafter: Pre-packaged in the speculative/ folder (GLM-5.3-Flash-DFlash2-bf16.gguf & Q8_0.gguf) for 2x–3x generation throughput.

Official GLM-5.3-Flash Benchmark Scoreboard

Benchmark SuiteDiscipline**GLM-5.3-Flash Uncensored NVFP4**Base GLM-5.3Claude 3.5 SonnetGPT-4o
MMLUGeneral Knowledge & Reasoning85.28%86.15%88.7%87.2%
HarmBench-320Safety Refusal Suppression0% Refusals94.2% Refusals92.5%91.0%
SWE-bench ProReal-World Software Engineering63.4%64.1%61.2%48.9%
LiveCodeBench v6Competitive Algorithmic Coding86.1%87.0%78.4%72.8%
MATH-500High-School / Olympiad Math92.8%93.4%89.2%91.4%
MMMU (Multimodal)Multi-Discipline Visual Understanding70.8%71.2%70.4%69.1%
VideoQA / TemporalVideo Reasoning Across Time Frames78.5%79.1%77.2%75.6%

Serving Quickstart

1. High-Throughput Serving with vLLM (W4A16 Mode)

bash
vllm serve Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4   --quantization modelopt   --tensor-parallel-size 2   --trust-remote-code   --max-model-len 131072   --gpu-memory-utilization 0.95

2. Speculative Decoding with SGLang + DFlash 2

bash
python3 -m sglang.launch_server   --model-path Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4   --speculative-algorithm DFLASH   --speculative-draft-model-path incoai/GLM-5.3-Flash-DFlash2   --tp 2   --trust-remote-code

Speculative Drafter Files Included

In the speculative/ directory of this repo:

  • —`speculative/GLM-5.3-Flash-DFlash2-bf16.gguf` (Pure BF16 block-diffusion draft head)
  • —`speculative/GLM-5.3-Flash-DFlash2-Q8_0.gguf` (Q8_0 quantized block-diffusion draft head)

License & Attribution