LLM EXPLORER 59,962 MODELS INDEXED

GLM 5.3 Flash MLX 4bit MTP by Vontra

By Vontra · 2953 downloads
HF Repo ↗ Copy URL

GLM 5.3 Flash MLX 4bit MTP is an open-source language model by Vontra. Features: 50.8b LLM, VRAM: 182GB, License: mit, Quantized, LLM Explorer Score: 0.49, ELO: 1286.

  Arxiv:2602.15763   4-bit   4bit   Apple-silicon Base model:quantized:zai-org/g... Base model:zai-org/glm-5.3-fla...   Conversational   En   Glm   Glm5   Glm5-next   Glm5 next   Image-text-to-text   Mixture-of-experts   Mlx   Mlx-vlm   Multimodal   Native-mtp   Omlx   Quantized   Region:us   Safetensors   Sharded   Speculative-decoding   Tensorflow   Vision   Vision-language   Zh
Model Card on HF 🤗: https://huggingface.co/Vontra/GLM-5.3-Flash-MLX-4bit-MTP 

GLM 5.3 Flash MLX 4bit MTP Benchmarks

LMSys ELO 1286 vs 1272 (so35) 1.1%
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

GLM 5.3 Flash MLX 4bit MTP Parameters and Internals

LLM NameGLM 5.3 Flash MLX 4bit MTP
Repository 🤗https://huggingface.co/Vontra/GLM-5.3-Flash-MLX-4bit-MTP 
Base Model(s)  zai-org/GLM-5.3-Flash   zai-org/GLM-5.3-Flash
Model Size50.8b
Required VRAM182 GB
Updated2026-08-30
MaintainerVontra
Model Typeglm5_next
Model Files  4.3 GB: 1-of-43   4.3 GB: 2-of-43   4.3 GB: 3-of-43   4.3 GB: 4-of-43   4.3 GB: 5-of-43   4.3 GB: 6-of-43   4.3 GB: 7-of-43   4.3 GB: 8-of-43   4.3 GB: 9-of-43   4.3 GB: 10-of-43   4.3 GB: 11-of-43   4.3 GB: 12-of-43   4.3 GB: 13-of-43   4.3 GB: 14-of-43   4.3 GB: 15-of-43   4.3 GB: 16-of-43   4.3 GB: 17-of-43   4.3 GB: 18-of-43   4.3 GB: 19-of-43   4.3 GB: 20-of-43   4.3 GB: 21-of-43   4.3 GB: 22-of-43   4.3 GB: 23-of-43   4.3 GB: 24-of-43   4.3 GB: 25-of-43   4.3 GB: 26-of-43   4.3 GB: 27-of-43   4.3 GB: 28-of-43   4.3 GB: 29-of-43   4.3 GB: 30-of-43   4.3 GB: 31-of-43   4.3 GB: 32-of-43   4.3 GB: 33-of-43   4.3 GB: 34-of-43   4.3 GB: 35-of-43   4.3 GB: 36-of-43   4.3 GB: 37-of-43   4.3 GB: 38-of-43   4.3 GB: 39-of-43   4.3 GB: 40-of-43   4.3 GB: 41-of-43   4.3 GB: 42-of-43   1.4 GB: 43-of-43
Supported Languagesen zh
Quantization Type4bit
Model ArchitectureGlm5NextForConditionalGeneration
Licensemit
Transformers Version5.16.0