---
title: "Discussion on \"There is no 4B Qwen 3.8. The smallest open model is 27B, and here is what it took to run it on one rented GPU.\" | Hashnode"
description: "Discussion on \"There is no 4B Qwen 3.8. The smallest open model is 27B, and here is what it took to run it on one rented GPU.\". Qwen 3.8 has no 2B, 3B or 4B model. I pulled qwen3.8:27b into Ollama on a rented NVIDIA L4: 17.7 GB on disk, 18,004 MiB of GPU memory in use, 18 to 35 tokens a second once loaded."
url: "https://hashnode.com/posts/there-is-no-4b-qwen-38-the-smallest-open-model-is-27b-and-here-is-what-it-took-to-run-it-on-one-rent/6ad03c7c28eb9de96cce725c"
---

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

Discussion

# [There is no 4B Qwen 3.8. The smallest open model is 27B, and here is what it took to run it on one rented GPU.](https://systemdesignacademy.hashnode.dev/there-is-no-4b-qwen-38-the-smallest-open-model-is-27b-and-here-is-what-it-took-to-run-it-on-one-rent?utm_source=hashnode&utm_medium=feed)

Qwen 3.8 has no 2B, 3B or 4B model. I pulled qwen3.8:27b into Ollama on a rented NVIDIA L4: 17.7 GB on disk, 18,004 MiB of GPU memory in use, 18 to 35 tokens a second once loaded.

Oct 15 · 12 min read · systemdesignacademy.hashnode.dev  
Read the article

-   [#qwen](https://hashnode.com/tag/qwen)
-   [#ollama](https://hashnode.com/tag/ollama)
-   [#llm](https://hashnode.com/tag/llm)
-   [#open-source](https://hashnode.com/tag/open-source)
-   [#gpu](https://hashnode.com/tag/gpu)

## Written by

[RONI DAS](https://hashnode.com/@ronidas)

@ronidas · India

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

AI engineering, agentic memory, machine learning, and system design at scale.

I work on AI systems and the infrastructure they actually run on: agentic architectures, memory for agents, retrieval, and the graph and data layers that keep model output grounded in something real. Formerly at Neo4j. I also write and teach the systems side that most AI content skips, from distributed systems and databases to the design decisions that decide whether a model ever reaches production. I run Total Technology Zonne on YouTube and I am building systemdesign.academy, an interactive course on system design, machine learning, and AI engineering, covering the parts most tutorials leave as an exercise for the reader. Mock interviews and prep packs: topmate.io/roni

## Responses 1

Markdown supported · ⌘↵ to post

[Damiano Bianda](https://hashnode.com/@damianobianda)

[Oct 17](https://hashnode.com/posts/there-is-no-4b-qwen-38-the-smallest-open-model-is-27b-and-here-is-what-it-took-to-run-it-on-one-rent/6ad03c7c28eb9de96cce725c/comment/6ad36d035000549f41aa2ffc)

Great teardown. The size gap you documented is real on the ground: when I needed a model for a fully offline tool (explaining official letters, the kind of text that cannot leave the machine), nothing in the newest families was small enough for plain CPU, so I ended up on a 1B Gemma. Slower than your L4 numbers, about 40 seconds per document end to end, but the task budget allowed it. Your memory math is the part most "run it locally" tutorials skip: 18 GB before context is not a laptop model for most people. Did you measure how much extra memory the context window adds as you raise it past 4,096?