Guanaco
The QLoRA demo chatbot that made single-GPU fine-tuning standard
Overview
Open research chatbots from the 2023 QLoRA paper, fine-tuned from LLaMA in 4-bit on a single GPU. The models are museum pieces now, but QLoRA itself became the standard method for cheap LLM fine-tuning across the open-source ecosystem.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
Use cases
ASEAN Perspective
Guanaco in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Guanaco existed to prove a point, and the point stuck: QLoRA's 4-bit quantised fine-tuning let a 65B LLaMA train on a single 48GB GPU, and the method is now baked into bitsandbytes, PEFT, Axolotl and essentially every open fine-tuning stack. The chatbots themselves — 7B to 65B, tuned on OASST1 — claimed 99.3% of ChatGPT on the Vicuna benchmark in May 2023, a number that says more about that benchmark than the model. There's no hosted service, no API, no support, and the LLaMA-1 base was never commercially licensed. Treat it as a landmark paper with demo weights attached: read the method, use QLoRA via your framework, skip the download.
What people say
99.3% of ChatGPT's quality from 24 hours on one GPU — that was Guanaco's May 2023 headline, and it deserves both the fame and the asterisk. The number came from GPT-4 judging outputs on the 80-prompt Vicuna benchmark, an evaluation method the field quickly learned to distrust. Nobody who used Guanaco 65B mistook it for ChatGPT. But the claim did its job: it got the world to read the QLoRA paper.
QLoRA was the actual product. Tim Dettmers and collaborators at the University of Washington showed you could quantise a base model to 4-bit (their NormalFloat data type), freeze it, and train small LoRA adapters on top — bringing a 65B fine-tune down to a single 48GB card and a 7B down to consumer hardware. That recipe is now everywhere: baked into bitsandbytes, Hugging Face PEFT, Axolotl, Unsloth and effectively every open fine-tuning stack. Few 2023 papers aged better.
The chatbots themselves aged like every other LLaMA-1 fine-tune, which is to say completely. The weights (7B–65B, tuned on the OASST1 dataset) still sit on Hugging Face, but no inference provider hosts them, there's no API or support, and the LLaMA-1 base was never licensed for commercial use anyway. A score of 4 is about right for what's listed: a landmark method demonstration wearing a chatbot costume. Read the paper, use QLoRA through your fine-tuning framework of choice, and don't download the model.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- Guanaco 65B was fine-tuned in 24 hours on a single GPU that costs $2/hour to rent — demonstrating that frontier-class model fine-tuning is accessible to independent researchers.
- The QLoRA technique developed for Guanaco reduces the memory required to fine-tune Llama 2 70B by 75% compared to full precision training.
- The QLoRA paper is one of the most cited machine learning papers of 2023 and directly enabled consumer-friendly fine-tuning tools like Axolotl and LLaMA Factory.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Guanaco's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Guanaco unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Guanaco directly →
Spotted something out of date? Suggest an update →
Guanaco in the news
Alternatives to Guanaco
More in LLMs & Chat
