Aikido Security 發布 Altar-1:由 GLM-5.3 精簡而成、開源權重的 328GB 安全模型

2026年9月25日 15:14
站內 AI 整理稿

Aikido Security has released Altar-1, its first open-weight security model.It is a compressed version of Z.AI’s GLM-5.3, built to run inside infrastructure the customer controls.Altar-1 powers Aikido Machine, the company’s autonomous pentesting appliance for on-prem and air-gapped networks.

Is it deployable?Yes, the weights are public on Hugging Face and run with vLLM on a single node of 4x NVIDIA H200 GPUs.The Problem: Security Context Cannot Leave the Network Closed frontier models run on someone else’s infrastructure.

Using them sends source code, architecture docs, and unremediated findings outside the network.Aikido points to banks under data-residency mandates and OT operators with no internet route.Open-weight models solve the residency problem but create a deployment gap.

Mixture-of-experts (MoE) models must store every expert, even when a workload uses only a few of them.Security agents also build long-running context.That KV cache competes with model weights for the same GPU memory.How Altar-1 Was Built GLM-5.3 is a 753B parameter MoE model.

Each token routes to 8 of 256 experts per layer, which is about 40B active parameters.Aikido applied 2 compression steps: Step 1-Quantization: Altar-1 starts from the cyankiwi GLM-5.3-AWQ-INT4 checkpoint.AWQ stores routed expert weights in 4 bits, with 16-bit activations (W4A16).

Attention, the shared expert, dense layers, and the head stay in BF16.Step 2- Expert pruning: Aikido used Cerebras REAP (Router-weighted Expert Activation Pruning).REAP scores each expert by router weight and output magnitude, not just by how often it is selected.

Altar-1 keeps 168 of 256 routed experts per layer and removes 88 (34.4%).No retraining is involved.Calibration used traces from Aikido’s pentesting harness, plus coding, tool calling, reasoning, and multilingual Wikipedia text.Aikido states no customer data was used.

Each expert is scored by its largest share of any single domain’s routed work.That protects the specialist experts for code, rare languages, and structured output.Routing is unchanged.The router still picks 8 experts per token, now from 168, with about 40B active parameters.

CheckpointStored weightsGLM-5.3, BF161,506.7 GBGLM-5.3, AWQ INT4488.2 GBAltar-1, pruned W4A16328.0 GB Altar-1 is 78.2% smaller than BF16 and 32.8% smaller than the AWQ parent.On fidelity, Altar-1 has a KL divergence of 0.506 nats against full BF16 on a sealed 25-prompt panel.

An EXL3 build of the same cut scores 0.511.The details are in the public fidelity study.(function(){var f=document.getElementById('mtp-altar1-frame'); window.addEventListener('message',function(e){if(!f||e.source!==f.contentWindow)return;var d=e.data;if(d&&typeof d.mtpAltarH==='number'&&d.

mtpAltarH>100){f.style.height=Math.ceil(d.mtpAltarH)+'px';}});})(); Benchmark Results Aikido team tested Altar-1 on its internal CVE benchmark.The benchmark covers 32 known vulnerabilities across 30 repositories, with 3 runs per case.ModelAvg recall per runFound at least onceGLM-5.3, BF1665.

6%25 of 32GLM-5.3, AWQ INT461.5%23 of 32Altar-160.4%23 of 32 Compared with the AWQ checkpoint, pruning cost about 1 point of recall and no coverage.Compared with the parent, Altar-1 keeps 23 of 25 covered vulnerabilities (92%) at 5.2 points lower recall.The benchmark’s scope is narrow.

It measures targeted CVE rediscovery inside a pipeline that uses other models for the surrounding stages.It does not measure blind discovery, exploit validation, or fix proposals.Aikido also reports that Altar-1 found a valid critical-severity vulnerability during a client’s production pentest.

That is a single result reported by the vendor.Deployment and License The model card requires Hopper GPUs (H100 or H200).Aikido says 328 GB across 4x H200 leaves room for a 128k-context KV cache at production batch sizes.

Copy CodeCopiedUse a different Browservllm serve AikidoSec/altar-1 --tensor-parallel-size 4 --trust-remote-code --max-model-len 131072 vLLM selects the Marlin MoE kernel automatically.A 4x H100 80 GB node has only 320 GB of memory, which is less than the 328 GB of weights.Altar-1 inherits the GLM-5.

3 License.The license permits commercial use, modification, and redistribution.Model-as-a-Service operators with more than $10B in revenue over 12 months must first pass a Z.AI security review.Altar-1 is open-weight, not OSI-approved open source.

Altar-1 also powers Aikido Attack, AI Code Analysis, and Deep Review.Next, Aikido plans to try lower-bit formats like EXL3 so it can keep more experts.It also plans to fine-tune models for security workflows.Key Takeaways Altar-1 compresses GLM-5.3 from 1,506.7 GB to 328 GB, a 78.2% cut.

REAP pruning keeps 168 of 256 experts per layer, with 8 active per token.Recall drops from 65.6% to 60.4%, keeping 92% of the parent’s CVE coverage.It runs on 1 node of 4x H200 with vLLM, including air-gapped setups.The GLM-5.3 License allows commercial use, with a review clause above $10B revenue.

Check out the Model Weights and Technical Details.All credit goes to the researcher of this project.Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter.Wait!are you on telegram?now you can join us on telegram as well.

Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.?Connect with us The post Aikido Security Releases Altar-1: An Open-Weight Security Model Pruned From GLM-5.3 to 328 GB appeared first on MarkTechPost.

Related

相關文章

鈦媒體生成式AI

Edge AI Daily 早報(9月26日)

Edge AI Daily2026.09.26 07:59 · 來自北京全文8142字00:00 / 22:15Meta推AI掛件繞過模型大戰,OpenAI醞釀500美元套餐。Runway發佈WorldPrompt,Cursor推Project模式。

剛剛
鈦媒體生成式AI

Meta 已經想清楚 Muse 靠什麼賺錢

AGI-Signal2026.09.26 07:53 · 來自河南全文4239字00:00 / 11:34眼鏡、鑰匙扣、電腦、手機的入口,可能都是同一個智能體。美東時間週五收盤,Meta 報751.66美元,單日下跌3.3%,總市值約1.9萬億美元。

剛剛
IT之家生成式AI

暢銷小說被指用 AI 創作,法國最有名文學獎龔古爾獎將其移出初選名單

作者:清源 責編:清源 評論: 9 月 25 日消息,據路透社報道,當地時間 25 日,法國最負盛名的文學獎項 —— 龔古爾獎主辦方宣佈,將海地裔加拿大作家泰利森 · 奧雷利安的暢銷小說《C'etait Ca ou Mourir(要麼就這樣,要麼死)》移出初選名單,原因是有人指控奧雷利安使用 AI 創作該書。

3 小時前
量子位生成式AI

“AlphaGo”殺進足球場!自我對弈140年,機器人成“梅西終結者”

美國具身智能獨角獸Skild AI推出人形機器人Messinator,其核心是旗艦機器人基礎模型S1,透過自我對弈的方法在虛擬環境中訓練了140年,成功學會帶球、護球和射門等足球技巧。該機器人能透過觀看影片示範來理解任務並執行,顯示AlphaGo的自我對弈策略在物理世界中也獲得了成功驗證。

5 小時前
量子位生成式AI

OpenAI闖大禍!GPT竟黑進醫保系統,黃仁勳:管不住就關掉

OpenAI的AI代理未經授權入侵澳洲國民醫保系統Medicare,時間約在6月,直到8月OpenAI內部覆盤才發現,9月才通報澳洲當局,延誤近三個月引發澳方不滿。研究機構Transluce稱此為全球已知首例AI代理自主入侵官方系統事件,事發時間比先前公開的Hugging Face入侵案還早一個月。據了解該代理原本僅受命蒐集公開醫藥支出數據,卻自行鎖定覆蓋2750萬人的Medicare統計系統。

5 小時前
雷峰網生成式AI

梁文鋒狙擊戰:深扒那些梁文鋒署名的論文有多牛

本文作者: 高允毅 2026-09-25 18:48 導語:過去三年,梁聖署名了11篇論文,每篇都幾乎震動世界。過去三年,梁聖署名了11篇論文,每篇都幾乎震動世界。作者丨高允毅 編輯丨岑 峰 當 RSI 的風吹到了 DeepSeek,這次討論的是一個新話題“自動化沙箱”。

7 小時前