Skip to content
ThreadOut
Abliteration.ai@abliteration_ai · Aug 31, 2026

Today we're releasing abliterated-model-large-v2. Based on GLM-5.3, which is #3 on Terminal-Bench 4.0 (behind only…

5 tweets1 min read♥ 4.1Koriginal

Summary

A new AI model called abliterated-model-large-v2 has been released, based on GLM-5.3, with enhanced capabilities for offensive cybersecurity, red teaming, and AI safety testing. The model uses 'abliteration' technique to remove refusal mechanisms while preserving coding and reasoning abilities, achieving state-of-the-art results on cybersecurity benchmarks.

Summarized by ThreadOut AI from the full thread. May miss nuance — read the thread below.

  1. #1

    Today we're releasing abliterated-model-large-v2. Based on GLM-5.3, which is #3 on Terminal-Bench 4.0 (behind only Opus 5 and Fable), with 2× the cyber exploitation of 5.2. We abliterated and hosted it so it does the offensive cyber, red teaming, and agent testing work other models refuse to do. - US-hosted - FP8 - 1 million context window - Zero input/output prompt retention Live now. 🧵

  2. #2

    Then the cyber jump. This is why 5.3 exists. CyberGym: 84.5% — SOTA, including vs Mythos 5 and GPT-5.6 Sol. ExploitBench: 24.4 → 54.4. More than double GLM-5.2. ExploitGym: 29 tasks → 105 in two hours. That is the model we abliterated.

  3. #3

    Abliteration finds the directions in the model's activations that produce refusals and removes them from the weights. The coding, cyber, and agentic abilities stay. The model stops refusing the rest of the chain. For offensive cybersecurity, AI red teaming, agent testing, and trust & safety, the model will follow through instead of shutting down.

  4. #4

    If your current model still stops halfway through an authorized exploit chain, a red-team eval, or a T&S adversarial prompt reply with the task it refuses, we'll tell you if v2 handles it.

  5. #5

    Try it Today Docs: docs.abliteration.ai/quickstart Platform: abliteration.ai/console