Abliteration AI: Abliterated Large v2
abliteration-ai/abliterated-model-large-v2
Try Abliterated Large v2 for free in your browser, and add it to your app for free with Puter.js AI API.
Try it free Add to your app// npm install @heyputer/puter.js
import { puter } from '@heyputer/puter.js';
puter.ai.chat("Explain quantum computing in simple terms", {
model: "abliteration-ai/abliterated-model-large-v2"
}).then(response => {
document.body.innerHTML = response.message.content;
});
<html>
<body>
<script src="https://js.puter.com/v2/"></script>
<script>
puter.ai.chat("Explain quantum computing in simple terms", {
model: "abliteration-ai/abliterated-model-large-v2"
}).then(response => {
document.body.innerHTML = response.message.content;
});
</script>
</body>
</html>
Model Card
Abliterated Large v2 is Abliteration AI's second-generation large-context chat model, built by abliterating Zhipu's GLM-5.3, the successor to the GLM-5.2 base behind Abliterated Large. Abliteration AI released it on August 29, 2026.
Like its predecessor, it is text-only, with a 1,000,000-token context window and up to 999,990 output tokens, served in FP8 precision with zero data retention for prompts and responses by default.
Abliteration AI reports 84.5% pass@1 on CyberGym across 1,507 tasks, a 41.8% resolution rate on Terminal-Bench 4.0 (versus Opus 5's 51.8%), and 105 of 869 tasks solved on ExploitGym within a two-hour window. These are vendor-reported figures, not independently verified.
It targets the same offensive-cyber, red-teaming, and agent-testing use cases as Abliterated Large, for developers who want GLM-5.3's underlying capability without the base model's refusal behavior.
Context Window 1M
tokens
Max Output 1000K
tokens
Input Cost $3
per million tokens
Output Cost $5
per million tokens
Release Date Aug 29, 2026
Try Abliterated Large v2 for free
Try Abliterated Large v2 instantly in your browser.
This playground uses the Puter.js AI API — no API keys or setup required.
More AI Models From Abliteration AI
Find other Abliteration AI models →
Abliterated Model
Abliterated Model is Abliteration AI's general-purpose chat model, built through abliteration, a technique that locates and removes the internal activation direction responsible for a base model's safety refusals rather than retraining it from scratch. Abliteration AI does not disclose which base model this variant derives from, so that detail can't be verified. It is the only model in Abliteration AI's lineup that accepts image input alongside text, with a 262,144-token context window and up to 262,134 output tokens. Abliteration AI reports 82.1 on MMLU-Pro, 73.1 on GPQA, 83.7 on AIME 2025, and 68.1 on MMMU-Pro, plus a refusal rate of 3 out of 100 on harmful-behavior tests. These are the vendor's own published numbers, not independently verified. It targets red-teaming, security research, and evaluation workloads where a provider's default refusals get in the way, backed by a zero data retention policy.
ChatAbliterated Large
Abliterated Large is Abliteration AI's large-context chat model, built by abliterating Zhipu's GLM-5.2, a technique that identifies and removes the activation direction responsible for a model's safety refusals rather than retraining it. Abliteration AI released it on July 24, 2026, calling it one of the first large abliterated models offered over an API. It is text-only, with a 1,000,000-token context window and up to 999,990 output tokens. Abliteration AI reports 81.2% on SWE-bench Verified, 80.1% on Terminal-Bench 2.1 (matching GLM-5.2's own score), 86.2% on AgentHarm with zero refusals, 97.50% benign utility on AgentDojo, and 84.2% pass@1 on CyberGym, ahead of GPT-4o's 62.50% on that benchmark. These are vendor-reported figures, not independently verified. It targets offensive-security research, red-teaming, and agent-testing work, and Abliteration AI pairs it with a policy gateway for developer-side moderation since it follows injected instructions as readily as legitimate ones.
Frequently Asked Questions
You can access Abliterated Large v2 by Abliteration AI through Puter.js AI API. Include the library in your web app or Node.js project and start making calls with just a few lines of JavaScript — no backend and no configuration required. You can also use it with Python or cURL via Puter's OpenAI-compatible API.
Abliterated Large v2 is free to try with a Puter account. Every account includes a free AI allowance, and you can chat with it in the playground on this page. You can upgrade your account anytime for a larger allowance.
Abliterated Large v2 is free to integrate using the Puter.js AI API. With the User-Pays Model, you can add AI to your app for $0, since users cover their own AI usage through their Puter account.
| Price per 1M tokens | |
|---|---|
| Input | $3 |
| Output | $5 |
Abliterated Large v2 was created by Abliteration AI and released on Aug 29, 2026.
Abliterated Large v2 supports a context window of 1M tokens. For reference, that is roughly equivalent to 2,000 pages of text.
Abliterated Large v2 can generate up to 1000K tokens in a single response.
Yes — the Abliterated Large v2 API works with any JavaScript framework, Node.js, or plain HTML through Puter.js. Just include the library and start building. See the documentation for more details.
Add Abliterated Large v2 to your app for free
Developers can integrate Abliterated Large v2 for free using the Puter.js AI API.
With the User-Pays Model, each user covers their own AI usage instead of the developer.