Enterprise Plans
AI Application Development
AI Translator API
AI SEO/GEO Service
Web Scraping Service
Log in
const { OpenAI } = require('openai');
const api = new OpenAI({
baseURL: 'https://api.ai.cc/v1',
apiKey: '',
});
const main = async () => {
const result = await api.chat.completions.create({
model: 'togethercomputer/guanaco-33b',
messages: [
{
role: 'system',
content: 'You are an AI assistant who knows everything.',
},
{
role: 'user',
content: 'Tell me, why is the sky blue?'
}
],
});
const message = result.choices[0].message.content;
console.log(`Assistant: ${message}`);
};
main();
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.ai.cc/v1",
api_key="",
)
response = client.chat.completions.create(
model="togethercomputer/guanaco-33b",
messages=[
{
"role": "system",
"content": "You are an AI assistant who knows everything.",
},
{
"role": "user",
"content": "Tell me, why is the sky blue?"
},
],
)
message = response.choices[0].message.content
print(f"Assistant: {message}")
-
AI Playground
Test all API models in the sandbox environment before you integrate.
We provide more than 300 models to integrate into your app.
Product Detail
✨ Guanaco-33B: A Powerful Open-Source LLM for Conversational AI
The Guanaco-33B model stands out as a high-quality, open-source large language model (LLM) designed for advanced chatbot applications. Developed by Tim Dettmers and released in April 2023, this 33-billion parameter model is a fine-tuned version of the LLaMA architecture. It achieves performance that is highly competitive with commercial chatbots such as ChatGPT, thanks to its innovative use of 4-bit QLoRA.
Key Model Details:
- • Model Name: Guanaco
- • Developer/Creator: Tim Dettmers
- • Release Date: April 2023
- • Version: 33B Parameters
- • Model Type: Text-based Large Language Model (LLM)
🚀 Core Features & Technical Innovations
Guanaco-33B incorporates several advanced features that contribute to its efficiency and high performance:
- • Multilingual Competence: Trained on the comprehensive OASST1 dataset, it excels in high-resource languages while supporting diverse linguistic inputs.
- • Efficient Adapter Use: Employs LoRA adapters with an r=64 configuration, seamlessly integrated across all layers of the base LLaMA model.
- • 4-bit QLoRA Fine-tuning: Utilizes 4-bit QLoRA with the NormalFloat4 datatype, optimizing the base model and adapters for reduced memory footprint and faster experimentation.
- • Lightweight Checkpoints: Offers adapter-only checkpoints, significantly lowering the barrier for local experimentation and development.
🧠 Architecture, Training & Data Insights
The Guanaco-33B model is built on the robust LLaMA architecture, a widely recognized Transformer-based language model. Its fine-tuning process relies on a key dataset:
Training Data:
- • Dataset Source: Finetuned exclusively on the OASST1 dataset, a collaborative, multilingual collection of open-source assistant conversations.
- • Scale & Scope: Comprises over 100,000 conversations in various languages, enabling diverse and open-ended dialogue capabilities.
- • Diversity & Bias: The multilingual nature of OASST1 helps mitigate inherent biases and enhances its ability to process varied inputs, though full dataset specifics are not public.
- • Knowledge Cutoff: The exact knowledge cutoff date for Guanaco-33B is not publicly disclosed, aligning with its continuous development as an open-source project.
💡 Intended Use, Ethical Guidelines & Licensing
Guanaco-33B is primarily developed for research purposes. It is essential for users to be aware that, like many experimental AI models, it may occasionally produce outputs that could be considered problematic.
Important Information:
- • Language Support: Optimal performance is observed in high-resource languages, reflecting the composition of its training data.
- • Ethical Framework: Anthropic has published ethical guidelines for its development and use, emphasizing transparency, accountability, and guarding against potential misuse.
- • License Details: The Guanaco-33B model operates under the Apache 2 License, allowing for broad commercial and non-commercial applications. However, access to the underlying LLaMA model weights requires separate licensing agreements.
📊 Performance Metrics & Benchmarking
Guanaco-33B has been rigorously evaluated across several benchmarks, including the prestigious Anthropic Chatbot Leaderboard. On these platforms, it consistently demonstrates strong competitive performance when compared to leading commercial AI models like ChatGPT and Google's BARD. It's noteworthy that its performance can vary based on specific language contexts and tasks not covered by the current evaluation benchmarks.
🖥️ API Integration Example
For developers looking to integrate Guanaco-33B into their applications, here's a standard API usage example:
<snippet data-name="open-ai.chat-completion" data-model="togethercomputer/guanaco-33b"></snippet>❓ Frequently Asked Questions (FAQs)
Q1: What is Guanaco-33B and what makes it unique?
Guanaco-33B is an open-source, 33-billion parameter text-based LLM developed by Tim Dettmers. It's unique for being fine-tuned from the LLaMA model using 4-bit QLoRA, offering competitive performance with commercial chatbots for research purposes.
Q2: What kind of data was Guanaco-33B trained on?
It was fine-tuned on the OASST1 dataset, which is a multilingual collection of over 100,000 open-source assistant conversations, enabling its broad conversational capabilities.
Q3: Can Guanaco-33B be used for commercial projects?
The Guanaco-33B model itself is available under the Apache 2 License, which permits both commercial and non-commercial use. However, using the underlying LLaMA model weights requires adherence to additional licensing requirements.
Q4: How does its performance compare to other chatbots?
Guanaco-33B has shown competitive performance against commercial chatbots like ChatGPT and BARD on various benchmarks, including the Anthropic Chatbot Leaderboard.
Q5: Are there any ethical considerations for using Guanaco-33B?
Yes, it's primarily for research, and users should be aware of potential problematic outputs. Anthropic has issued ethical guidelines focusing on transparency, accountability, and preventing misuse, which should be followed.
Learn how you can transformyour company with AICC APIs
