Unsloth Dynamic GGUFs
Hacker ApprovedQuantized model in AI tools
Compressed GGUF builds of popular open models that run well on consumer hardware, with careful quality tuning. Caveat: critics say releases lack version numbers and early quants had bugs.
Compressed GGUF builds of popular open models that run well on consumer hardware, with careful quality tuning. Caveat: critics say releases lack version numbers and early quants had bugs.
22 people gave an opinion
- recommend it
- 15
- criticize it
- 7
- approval share
- 68%
Recommended in 4 of 13 months
- November 2025: not recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: recommended
- March 2026: not recommended
- April 2026: recommended
- May 2026: not recommended
- June 2026: recommended
- July 2026: not recommended
- August 2026: recommended
- September 2026: not recommended
- October 2026: not recommended
- November 2026: not recommended (month in progress)
Why it carries the badge
- 15 people recommend itNeeds 10 or more, met.
- 2.1 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 4 of 13 monthsNeeds 4 or more, met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Good quality at low bit widths
“not all layers can be 1bit - important ones are in 8bit or 16bit, and we show it still works well”
Comment from 16 Apr 2026 in “Qwen3.6-35B-A3B: Agentic Coding Power, Now Open to All” “So q6 is practically perfect, and q3 is meaningfully decent.”
Comment from 28 Feb 2026 in “Unsloth Dynamic 2.0 GGUFs” “they seem to work remarkably well”
Comment from 28 Feb 2026 in “Unsloth Dynamic 2.0 GGUFs” Runs well on consumer hardware
“I would suggest using UD-IQ3_XXS for 10.9GB for 16GB machines”
Comment from 19 Aug 2026 in “Unsloth Dynamic 3.0 GGUFs” “Your stuff ran LLMs extremely well on < $500 boxes”
Comment from 28 Feb 2026 in “Unsloth Dynamic 2.0 GGUFs” “working pretty well so far on 4090 16gb”
Comment from 19 Aug 2026 in “Unsloth Dynamic 3.0 GGUFs” Fixes bugs and shares research
“Your original Q4_K_XL quant was broken. Fixing bugs is great. I use your quants and appreciate your work.”
Comment from 28 Feb 2026 in “Unsloth Dynamic 2.0 GGUFs” “which shows SOTA on nearly all bits, and we shared all our research as well”
Comment from 28 Feb 2026 in “Unsloth Dynamic 2.0 GGUFs” Go-to source for GGUF downloads
“your gguf are the first ones I look for when I want to download a gguf model”
Comment from 19 Aug 2026 in “Unsloth Dynamic 3.0 GGUFs” “Very impressed by unsloth's team releasing the GGUF so quickly”
Comment from 2 Apr 2026 in “Google releases Gemma 4 open models”
What people warn about
Each mark is one comment
Early quant bugs and evasive replies
“Your original Q4_K_XL quant was broken. Fixing bugs is great. I use your quants and appreciate your work.”
Comment from 28 Feb 2026 in “Unsloth Dynamic 2.0 GGUFs” “This isn't a postmortem, it's PR fluff without actually addressing the issue.”
Comment from 28 Feb 2026 in “Unsloth Dynamic 2.0 GGUFs” “This is a famous non-apology non-explanation of what actually happened.”
Comment from 3 Mar 2026 in “Unsloth Dynamic 2.0 GGUFs” - Comment from 28 Feb 2026 in “Unsloth Dynamic 2.0 GGUFs”
Unclear versioning and benchmark baselines
“But those were compared against UD2.0 I guess.”
Comment from 20 Aug 2026 in “Unsloth Dynamic 3.0 GGUFs” “It would be nice if unsloth published GGUFs would use a version number or something”
Comment from 19 Aug 2026 in “Unsloth Dynamic 3.0 GGUFs”
What people recommend instead
The same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| Unsloth GGUF quants | 35 | 10 | 78% | 10 | — | Approved |
| Unsloth Dynamic GGUFsThis page | 15 | 7 | 68% | 4 | — | Approved |
What people said
27 of the 29 comments that mention it, newest first. Each links to the original comment.
“3.8 is a significant improvement on 3.6.”
“it look like UD3.0 Q3_K_XL based on the Unsloth blog post”
“But those were compared against UD2.0 I guess.”
“For some reason the unsloth models leave hardly any room for context.”
“It would be nice if unsloth published GGUFs would use a version number or something”
“I would suggest using UD-IQ3_XXS for 10.9GB for 16GB machines”
“working pretty well so far on 4090 16gb”
“retain around 72% top-1% accuracy yet being 89% smaller”
“your gguf are the first ones I look for when I want to download a gguf model”
“Unsloth is notorious for this”
“not all layers can be 1bit - important ones are in 8bit or 16bit, and we show it still works well”
“Very impressed by unsloth's team releasing the GGUF so quickly”
“The ggufs from unsloth don't work with ollama.”
“This is a famous non-apology non-explanation of what actually happened.”
“MXFP4 is much worse on many tensors”
“This isn't a postmortem, it's PR fluff without actually addressing the issue.”
“Don't bother with anything else.”
“Your stuff ran LLMs extremely well on < $500 boxes”
“Your original Q4_K_XL quant was broken. Fixing bugs is great. I use your quants and appreciate your work.”
“they seem to work remarkably well”
“Q8_K_XL looks mighty nice.”
“which shows SOTA on nearly all bits, and we shared all our research as well”
“No our Qwen3.5 new ones show the opposite”
“So q6 is practically perfect, and q3 is meaningfully decent.”
“there doesn't appear to be meaningful difference between their quants and bartowsky quants”
“they messed up the first version of their quants for Qwen 3.5”
“ours are dynamically quantized and calibrated with a reasonably large diverse dataset”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.