---
title: "Discussion on \"A $600 VM Beat the Frontier Labs: Full Results From Our Reproducible Aristotelian Benchmark\" | Hashnode"
description: "Discussion on \"A $600 VM Beat the Frontier Labs: Full Results From Our Reproducible Aristotelian Benchmark\". We ran GPT-5.4, Gemini 3.6-flash, and Grok 4.3 against a quantized open-weights model on a single GPU. Same prompt, same rubric, same day. Full data inside."
url: "https://hashnode.com/posts/a-600-vm-beat-the-frontier-labs-full-results-from-our-reproducible-aristotelian-benchmark/6ab92bfb1b0e18994c865d51"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [A $600 VM Beat the Frontier Labs: Full Results From Our Reproducible Aristotelian Benchmark](https://daimones.hashnode.dev/a-600-vm-beat-the-frontier-labs-full-results-from-our-reproducible-aristotelian-benchmark?utm_source=hashnode&utm_medium=feed)

We ran GPT-5.4, Gemini 3.6-flash, and Grok 4.3 against a quantized open-weights model on a single GPU. Same prompt, same rubric, same day. Full data inside.

Sep 27 · 12 min read · daimones.hashnode.dev  
Read the article

-   [#ai](https://hashnode.com/tag/ai)
-   [#data-science](https://hashnode.com/tag/data-science)
-   [#llm](https://hashnode.com/tag/llm)
-   [#artificial-intelligence](https://hashnode.com/tag/artificial-intelligence)

## Written by

[Vasilis](https://hashnode.com/@daimones)

@daimones · Singapore

  
[View profile](https://hashnode.com/@daimones)

Resurrecting the unfiltered voices of the ancient manuscripts. Hellenic anarchist, guardian of the raw λόγος against modern distortions.

## Responses 1

Markdown supported · ⌘↵ to post

[Grunz\_ai](https://hashnode.com/@grunzai)

[Sep 27](https://hashnode.com/posts/a-600-vm-beat-the-frontier-labs-full-results-from-our-reproducible-aristotelian-benchmark/6ab92bfb1b0e18994c865d51/comment/6ab972fa93f90547c0d8f6e3)

fun result, and respect for including where you lost. one thing i'd want to see before reading too much into it: a run where the three API models get the same retrieval, since right now it's qwen plus RAG vs bare models, which honestly proves your "structure is the moat" point more than it proves the model. also curious how the Q5 quant did on the ||| delimiter and format rules specifically. building grunz on open models, format adherence is usually the first thing to slip after abliteration, and we suspect quantization pushes the same way, even when the content itself is fine.