Sakhanda Wire
NVDA $154.93 -0.77% MSFT $505.37 -2.44% GOOGL $279.95 -1.33% META $618.13 -1.32% AMZN $257.42 -0.13%
← Back to the news

Aikido Security releases 328 GB open-weight security model

Aikido Security has released Altar-1, its first open-weight security model, designed to operate within infrastructure controlled by customers. The model is a compressed version of Z.AI’s GLM-5.3 and powers Aikido Machine, an autonomous penetration-testing appliance intended for on-premises and air-gapped networks. Its weights are available on Hugging Face and can run with vLLM on one node equipped with four NVIDIA H200 GPUs.

GLM-5.3 has 753 billion parameters and routes each token to eight of 256 experts. Aikido began with an AWQ INT4 checkpoint, then used Cerebras REAP to remove 88 experts per layer without retraining. Altar-1 retains 168 experts, while keeping about 40 billion active parameters per token. The resulting checkpoint occupies 328 GB, compared with 1,506.7 GB for the BF16 model and 488.2 GB for its AWQ parent.

On Aikido’s internal benchmark covering 32 vulnerabilities in 30 repositories, Altar-1 achieved 60.4% average recall per run and found 23 vulnerabilities at least once. The parent BF16 model reached 65.6%. Aikido says the model requires Hopper GPUs and inherits the GLM-5.3 license, which permits commercial use but includes a Z.AI review requirement for certain high-revenue model-as-a-service operators. It is open-weight, not OSI-approved open source.

Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.

Originally published by MarkTechPost on

Read the original on MarkTechPost ↗

The reporting and the image belong to MarkTechPost and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news