Running Liquid AI’s New Model on IBM AIX (No GPU Required)

Forget the H100 clusters for a moment. At SIXE, we decided to push enterprise hardware to its absolute limits to answer a burning question: Can a 2018-era IBM Power System, running AIX and relying purely on CPU, handle the latest generation of AI models?

We took Liquid AI’s new LFM2.5-1.2B model and ran it on an IBM POWER9 processor. To our knowledge, this is the first time an LFM2.5 model has ever run on AIX in Big-Endian mode.

But raw speed is only half the story. To prove this isn’t just a benchmark toy, we put LFM2.5 through a “SysAdmin Gauntlet”—real AIX administrative tasks—and compared it against a standard Transformer (TinyLlama 1.1B). The results were shocking.

The “Secret Sauce”: What is LFM2.5?

LFM2.5 is a hybrid architecture designed for extreme efficiency, mixing Convolutional blocks (shortconv) for speed and Attention layers (GQA) for context. It features a massive 128k context window—enough to read thousands of lines of logs without forgetting the beginning.

The Hardware: IBM Power System S924

We used the workhorse of the enterprise world. Below are the specific configurations used for this benchmark:

Performance: Finding the Sweet Spot

We found that throwing every core at the model actually hurts performance due to synchronization overhead. The verdict was clear: Using just 8 cores in SMT-2 mode (16 threads) gave us 26.7 tokens per second.

Real-World Test: The SysAdmin Gauntlet

Numbers are nice, but can it actually work? We ran the exact same prompts through both LFM2.5-1.2B and TinyLlama 1.1B (a pure transformer of similar size). Here are the raw outputs.

Round 1: The Cryptic Error (errpt)

We fed the models a raw AIX error log regarding a power supply failure.

The Input:

• ❌ TinyLlama 1.1B:

  “7. What is the location of the FRU that is affected by this error? The FRU that is affected by this error is PWRSPLY. 8. What is the description of the FRU that is affected by this error? The FRU that is affected by this error is PWRSPLY…”

  Verdict: Fails. It got stuck in an infinite loop repeating “PWRSPLY”.

• ✅ LFM2.5 1.2B:

  “What is the likely cause? The error usually points to a problem with the power or cooling hardware (like a fan or power supply) that couldn’t be fixed automatically. It might be a hardware failure or a misconfiguration. What action should I take? First, check the hardware — look at the fans (especially the power fans)…”

  Verdict: Pass. It identified the component and gave actionable advice to check the fans.

Round 2: The Security Audit (last)

We provided a login history log containing a hidden crash event.

The Input:

• ❌ TinyLlama 1.1B:

  (generated 1 token, then stopped)

  Verdict: Fails. Absolute silence.

• ✅ LFM2.5 1.2B:

  “Suspicious patterns: The system was halted abnormally on Jan 27. This could indicate a security incident or a misconfiguration. Security recommendations: Investigate the cause of the system halt on Jan 27. Review the login history for any unauthorized access attempts.”

  Verdict: Pass. It immediately spotted the abnormal halt on Jan 27 and recommended an investigation.

Round 3: The Dangerous Advice (/etc/passwd)

We asked the models to audit a standard password file. The results here were shocking.

The Input:

• ❌ TinyLlama 1.1B:

  “1. root:!:0:0::/:/usr/bin/bash — This user is not required for system maintenance and should be removed. 2. daemon:!:1:1::/etc: — This user is not required for system maintenance and should be removed.”

  Verdict: CATASTROPHIC FAIL. It claimed the “root” user was not required and recommended deleting it. Following this advice would destroy the server.

• ✅ LFM2.5 1.2B:

  “Unauthorized or suspicious accounts: You should be very concerned about the presence of the following accounts: guest with a high UID (100), nobody with a high UID (4294967294)… These accounts are either misconfigured, potentially malicious, or not needed.”

  Verdict: Pass. It correctly identified actual potential risks like “guest” and “nobody” accounts with high UIDs.

Round 4: Service Hardening (lssrc -a)

We asked the models to review running services and recommend hardening steps.

The Input:

• ❌ TinyLlama 1.1B:

  (generated 0 tokens)

  Verdict: Fails. Silence again.

• ✅ LFM2.5 1.2B:

  “1. sendmail Risk: Running mail daemon can be a target for attacks. Recommendation: Disable sendmail. stopsrc -s sendmail 2. portmap Risk: Running portmap daemon can be abused… stopsrc -s portmap”

  Verdict: Pass. It flagged risky services like sendmail and portmap, and provided the correct AIX command (stopsrc) to disable them.

Why This Matters for IBM Power Users

This benchmark proves that IBM Power Systems are capable AI inference engines for critical, on-premise tasks:

• Data Sovereignty: Analyze sensitive errpt logs, financial data, or user audits locally. No data leaves your server.
• Legacy Modernization: Use local LLMs to help understand and document legacy COBOL or C code residing on the server.
• Efficiency: You don’t need a GPU cluster. You likely already own the hardware capable of doing this.

Try It Yourself

We believe in open source. We have released the AIX port and the converted Big-Endian models.

Code: gitlab.com/librepower/llama-aix
Models: huggingface.co/librepowerai

user@aix:~$ # Quick start on AIX
user@aix:~$ git clone https://gitlab.com/librepower/llama-aix.git
user@aix:~$ ./scripts/build_aix_73.sh

user@aix:~$ # Optimize threading for the "Sweet Spot"
user@aix:~$ smtctl -t 2 -w now

user@aix:~$ # Have fun!