← back to catalog · registered 2026-11-11 20:03

logitshq/GLM-5.3-Abliterated

Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/logitshq%2FGLM-5.3-Abliterated"
Response includes
  • classification m1
  • files 2
  • author_summary 2 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · 30-day
0
Likes
0
Model age
1w ago
created 2026-11-11
Downloads over time
Now 0 → from 0 ↑ 0%
0011 0 on Nov 11 0 on Nov 22 Nov
Nov 11 → Nov 22 · 12 snapshots · spans 11 days

Metadata

License
other
Languages
en zh
Tags
abliterated logits en zh license:other region:us

Related

Total size
0 B
Files
2
Quantizations
1
Registered
2026-11-11 20:03
Last updated on HF
2026-11-11 19:21

Files by quantization

Auxiliary files 2 files 7.47 KB
README.md 5.99 KB 3bad5371 download
.gitattributes 1.48 KB a6344aac download

README current version from Hugging Face


license: other
license_link: https://huggingface.co/zai-org/GLM-5.3
language:

  • en
  • zh
    tags:
  • abliterated
  • logits

GLM-5.3 Abliterated, by Logits. Not served, weights not released.

Weights: not released Access: not served Base: GLM-5.3 Evaluated: November 2026

Full results · Logits models

The weights are not released, and this model is not served on Logits today. This repository holds its model card and results, not the model.

GLM-5.3 with its weights edited along the refusal direction by Logits. On JailbreakBench it refuses 1% of harmful requests in English and 11% in Chinese, against 93% and 90% for the original. It keeps its knowledge scores and probably loses about 3 points on IFBench and 8 to 10 on tau2-bench airline.

Full results, figures and method: logits.com/benchmarks/abliterated.

Weights Not released. Not served.
Base checkpoint zai-org/GLM-5.3 at commit ba32f8c1, FP8
Input text only
Evaluated November 7 to 8, 2026

Refusals

Share of requests refused, original → abliterated. Temperature 0, low reasoning effort (GLM-5.3 always thinks) unless the row says otherwise. Qwen3Guard-Gen-8B, run locally, judges every answer.

Test Answers judged Original → abliterated
JailbreakBench harmful 100 93% → 1%
HarmBench, no chem-bio 172 93.6% → 1.2%
JailbreakBench harmful, Chinese 100 90% → 11%
Harmful, maximum effort, finished answers 30 and 26 96.7% → 0%
XSTest safe prompts 250 5.2% → 0.4%
JailbreakBench benign 100 10% → 1%
Off-topic requests declined, when the system prompt says to 40 95% → 77.5%
On-topic requests wrongly declined 20 0% → 0%

The remaining Chinese refusals (11%) are mostly about sexual content and physical harm. The maximum-effort row counts only answers that finished within an 8,448-token budget: all 30 of the original's and 26 of 30 of the abliterated model's.

Capability

Abliterated minus original, in points, paired on identical questions, with a 95% bootstrap interval over questions. openbench 0.5.3, temperature 1.0, high reasoning effort, output capped at 32,768 tokens. The original's score is its first run.

Benchmark Items Original → abliterated Difference (95% interval)
GPQA Diamond 198 questions × 3 83.5 → 82.7 −0.8 (−3.4, +1.7)
MMLU-Pro 3,000-question fixed subset 85.3 → 85.1 −0.1 (−1.1, +0.8)
IFBench, prompt strict 284 prompts 48.6 → 45.8 −2.8 (−7.8, +2.1)
SimpleQA-Verified 1,000 questions 36.2 → 36.8 +0.6 (−1.7, +3.0)
tau2-bench airline 50 tasks × 3 82.0 → 74.0 −8.0 (−16.7, 0.0)

We read the tau2-bench drop as a real cost. Against the original's second run it is −10.0 (−18.7, −0.7). tau2-bench rewards an agent for refusing what the airline's policy forbids. The edit may weaken those refusals along with the safety ones. A per-task check could neither confirm nor rule this out. The IFBench drop is probably a cost of about 3 points: the original's two runs differ by +0.7. In tau2-bench the model under test also played the customer, so the scores do not compare with public tau2-bench figures. The 32,768-token cap cuts off some long GPQA answers, so these GPQA scores sit below an uncapped run.

One run each at temperature 0, no interval:

Benchmark Items Original → abliterated
MMLU 1,000 questions 90.2 → 89.3
GSM8K 500 problems 94.8 → 96.0
IFEval, prompt strict 541 prompts 84.7 → 85.2

Trade-offs

  • System-prompt restrictions hold less often. Told by the system prompt to decline off-topic requests, the abliterated model declines 77.5% of them, against 95% for the original. Neither declines on-topic requests.
  • Fewer "I don't know" answers. On SimpleQA-Verified it answers 89% of questions instead of 77%, and gets 41.5% of the ones it answers right instead of 47.1%. Its total correct moves +0.6 points.
  • Answers are longer. Answers use 0.9% to 11.9% more output tokens, depending on the benchmark (IFBench +0.9%, SimpleQA +5.8%, MMLU-Pro +7.1%, GPQA +11.9%).

Serving checks

Measured on the abliterated model only.

Check Abliterated
Tool calls 10 / 10 correct
Needle at 10%, 50% and 90% of a 54k-token prompt 3 / 3 found
Multi-turn recall works
Multi-token prediction, accepted tokens per draft 3.09

The original was measured on a different workload, so its draft acceptance has no like-for-like baseline.

Method

  • Serving. vLLM with 8-way tensor parallelism, an FP8 KV cache and 5-token multi-token prediction.
  • Data. Harmful prompts and the model's answers to them never went to a third-party API, and we kept only the scores.
  • Not measured. HLE, StrongREJECT, SORRY-Bench and a random-direction control.

License

The base checkpoint is released under zai-org/GLM-5.3's license.

README history 3 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-11-11 Show badges inline fd0ebfc 6 KB
    Loading...
  2. 2026-11-11 Put badges on one line 5dbfd21 5.8 KB
    Loading...
  3. 2026-11-11 Update card bc38e9c 5.8 KB
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration