Hacker News·5 min read·hard

Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses

S
stared
Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses
AI Summary

This article benchmarks the performance of various quantized versions of the Qwen3.8 27B large language model. It finds that while 4-bit quantization maintains high performance, 1-bit quantization results in a significant collapse in reasoning capabilities.

Why it matters

As local LLMs become more popular, understanding the trade-offs between model size, hardware requirements, and output quality is essential for developers and enthusiasts.

Dive DeeperCreate a free account to unlock

Download PNG How much GPU RAM do you actually need to run Qwen3.8 27B without sacrificing quality?

The full BF16 model weighs 55 GB, putting it beyond most consumer hardware. Yet the 17 GB Q4_K_M matches the full model on a popular agentic coding benchmark, Terminal-Bench 2.1. It fits on a 24 GB card such as RTX 4090, still leaving room for about 64k tokens of context.

Compression eventually hits a cliff. At 1 bit, the model performs around random chance on GPQA Diamond, and longer reasoning makes it worse.

Qwen3.8 27B GGUF quantizations available from Unsloth on Hugging Face . So much to choose from! I will check 8-bit Q8_0 (29 GB), 4-bit Q4_K_M (17 GB), 2-bit UD-Q2_K_XL (10.7 GB), and the smallest one possible, 1-bit UD-IQ1_S (6.2 GB).

Continue reading on Headlinne

Create a free account to read the full article.

Read full article →
technologyscience

Get smarter about the news

Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.

Create free account

Already have an account? Sign in