OpenAI’s new GPT 3.5 Instruct

GPT-3.5-Turbo-Instruct

OpenAI has unveiled a new and advanced model, “gpt-3.5-turbo-instruct“, designed to seamlessly interpret and execute instructions. The new model, which is now the default language models accessible on our API, is engineered to provide coherent and contextually relevant responses, making it a versatile asset for a range of applications. In this article, we’ll explore the functionalities and distinctive features of gpt-3.5-turbo-instruct and discuss why OpenAI embarked on the development of this model.

Why OpenAI Developed gpt-3.5-turbo-instruct:

According to openAI, the release of gpt-3.5-turbo-instruct is a big leap to improve users interaction with its models. It was trained to address problems older models had, giving clearer and more on-point answers. This makes it a good fit for all kinds of uses, whether you know a lot about tech or not.

What’s new?

gpt-3.5-turbo-instruct model is a refined version of the GPT-3, designed to perform natural language tasks with heightened accuracy and reduced toxicity. The GPT-3 models, while revolutionary, had a propensity to generate outputs that could be untruthful or harmful, reflecting the vast and varied nature of their training data sourced from the Internet.

Difference between GPT-3 and gpt-3.5-turbo-instruct models

To mitigate these issues and align the models more closely with user needs, OpenAI has employed reinforcement learning from human feedback (RLHF), a technique involving real-world demonstrations and evaluations by human labelers. This approach has enabled the fine-tuning of the model, making it more adept at following instructions and reducing the generation of incorrect or harmful outputs.

GPT-3 vs GPT-3.5 Turbo Instruct

The gpt-3.5-turbo-instruct diverges from its predecessor, GPT-3.5, in its core functionality. It is not designed to simulate conversations but is rather fine-tuned to excel in providing direct answers to queries or completing text. OpenAI asserts that this model maintains the speed efficiency synonymous with GPT-3.5-turbo.

Model Name	Use Cases	Advantages	Best for	Max Tokens
gpt-3.5-turbo	Natural language or code generation	Most capable and cost-effective, optimized for chat, receives regular updates	Traditional completions & chat interactions	4,097
gpt-3.5-turbo-16k	Natural language or code generation	Offers 4 times the context compared to the standard model	Scenarios requiring extended context	16,385
gpt-3.5-turbo-instruct	Natural language or code generation	Compatible with legacy Completions endpoint, similar capabilities as text-davinci-003	Instruction-following tasks	4,097
gpt-3.5-turbo-0613	Natural language or code generation	Includes function calling data, snapshot of gpt-3.5-turbo from June 13th 2023	Function calling data needs	4,097
gpt-3.5-turbo-16k-0613	Natural language or code generation	Snapshot of gpt-3.5-turbo-16k from June 13th 2023	Scenarios requiring extended context	16,385
gpt-3.5-turbo-0301	Natural language or code generation	Snapshot of gpt-3.5-turbo from March 1st 2023	–	4,097
text-davinci-003	Any language task	High-quality, longer output, consistent instruction-following, supports additional features	Diverse language tasks	4,097
text-davinci-002	Any language task	Trained with supervised fine-tuning	–	4,097
code-davinci-002	Code-completion tasks	Optimized for code-completion tasks	Code-completion tasks	8,001

GPT-3 vs GPT Instruct

The table above provides a concise overview of various models developed by OpenAI, each with unique capabilities and optimizations. It outlines the specific use cases, advantages, and maximum tokens for each model, offering insights into their functionalities and optimal applications. The models range from those optimized for natural language or code generation, like gpt-3.5-turbo and its variants, to those specialized in diverse language tasks and code-completion tasks, like text-davinci-003 and code-davinci-002.

How to Use gpt-3.5-turbo-instruct Model with Python

gpt-3.5-turbo-instruct is a completion model, hence you will need to use the completion function to get responses.

To interact with gpt-3.5-turbo-instruct using Python, you can refer to the following simplified code snippet.

Install the openai pip library

pip install openai

Import openai in your Python file

import openai

openai.api_key = "sk......." #Your openai API key
prompt = "Explain the concept of infinite universe to a 5th grader in a few sentences"

OPENA_AI_MODEL = "gpt-3.5-turbo-instruct"
DEFAULT_TEMPERATURE = 1

response = openai.Completion.create(
model=OPENA_AI_MODEL,
prompt=prompt,
temperature=DEFAULT_TEMPERATURE,
max_tokens=500,
n=1,
stop=None,
presence_penalty=0,
frequency_penalty=0.1,
)   

print(response["choices"][0]["text"])

InstructGPT-3.5 answer is below:

An infinite universe means that the universe is never-ending and has no boundaries. It keeps going and going, and we will never reach the end of it no matter how far we travel. Just like numbers go on forever, the universe goes on forever too. It's like a huge never-ending playground of planets, stars, and galaxies.

At Next Idea Tech, we are dedicated to exploring the frontier of the latest technologies and AI advancements, such as OpenAI’s, to propel businesses into a future of seamless automation and enhanced workflows. We leverage the expertise of our talented LATAM software engineers to harness the power of cutting-edge AI and tailor solutions that drive efficiency and innovation in your business processes.

Whether you are looking to automate intricate tasks or improve existing workflows, our team of experienced software developers is here to turn your visions into reality. Don’t hesitate to reach out and discuss your next project with us.

Zak Elmeghni