Agents / Huihui · Z.ai base model

GLM-5.3 Flash Abliterated

Huihui’s abliterated GGUF of Z.ai’s GLM-5.3-Flash, a 321B mixture-of-experts model for coding and agents. Builds from 93 GB to 292 GB for machines with 128–256 GB of memory and up.

Verified sourceReleased Sep 26, 2026Source checked 11/17/2026
LOCALOPEN WEIGHTS
Freshness 81 out of 100. Release age: 56 days old, Sep 26, 2026. Measures age, not quality.
Popularity 64 out of 100. 44,043 downloads in 30 days; 438 lifetime likes. Checked Nov 21, 2026. Measures adoption, not quality.
What do these scores mean?

Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.

Freshness · release age

A new release starts at 100. The score halves every 180 days: 100 × 2^(−age in days / 180). When the release date is unavailable, a repository date is a fallback—not proof of when the model was released. Missing dates stay unscored.

Check the date source
Popularity · adoption

Downloads over the past 30 days carry 80% of the score; lifetime likes carry 20%. Both use a logarithmic scale, capped at 10 million downloads and 100,000 likes. If likes are unavailable, downloads carry all the weight. Missing download statistics stay unscored.

round(80 × min(1, log10(1 + downloads) / 7) + 20 × min(1, log10(1 + likes) / 5))

Check the statistics source
Scores as of Nov 21, 2026. Statistics checked Nov 21, 2026. Statistics are marked for refresh after 72 hours.Full scoring method
WORKFLOWS FOR THIS MODEL 1

What do you want to do?

Choose a task. Get the software, files, settings, and steps to run it.

MAKE IT PERSONALChoose your hardware

Open a guide for its software, hardware requirements, and setup steps.

At a glance

Parameters
321B total · about 18B active per token
Architecture
Mixture of experts (glm5-next) · text and image in, text out
Context length
Huihui’s command uses 262,144 · the agent guide starts at 65,536
License
MIT
Software
llama.cpp
MY HARDWARE

Will it run on your machine?

Save your machine to see a personalized rating and its reasoning.

Add my hardware

Best for

  • Run a local Hermes agent that reads and writes files with your approval.
  • Private coding and long tasks with fewer refusals on a 128–256 GB Mac Studio or workstation.
  • Ask about images by loading the vision file alongside the model.

Tradeoffs

  • Huihui ablated only layers 15 to 35 and left the expert modules untouched, and its card reports no refusal test, so some refusals may remain.
  • The files are abliterated copies of Unsloth’s GLM-5.3-Flash GGUFs, not a new quantization from full precision.
  • Downloads are 93–292 GB and split into several parts.
  • A 2-bit or 3-bit build of a model this size loses precision; no quality measurements are published for these files.
  • A different model from the dealignai GLM-5.3 Uncensored page on YouRunAI, which covers the larger GLM-5.3.

Ways to run it

llama.cpp · Windows, macOS, Linux

The tag from Huihui’s model page. Used by the Hermes agent guide on this page, which also offers the smaller UD-Q2_K_XL and UD-Q3_K_XL builds.

llama-server -hf huihui-ai/Huihui-GLM-5.3-Flash-abliterated-GGUF:UD-Q4_K_XL --jinja -c 65536
Official instructions

What to expect

The pick for 128–256 GB of memory. On a 128 GB machine use UD-Q2_K_XL and expect a tight fit; with 256 GB use UD-Q4_K_XL.

Explore its uses

UNCENSOREDABLITERATEDAGENTSCODINGMIXTURE OF EXPERTSVISIONGGUFGLM-5.3
REFERENCE LIBRARY

Sources and files

Primary sources and files behind this model page.

4 SOURCES

Keep exploring