The Essential List: AI Text-Generation Models and Apps

Everything you need to know about text-generation models and apps

Ben Wodecki, Jr. Editor

July 19, 2023

11 Min Read
image of a computer head profile and computer code across a computer screen
iambuff/Getty Images

Artificial intelligence is giving people the power to create at scale and businesses the capability to work smarter and more efficiently.

One promising area already making an impact is natural language text generation. Powerful machine learning models such as GPT-4 and PaLM 2 can generate human-like text from simple prompts. Such systems have the potential to automate a wide range of tasks to increase productivity and yield insights more quickly.

AI Business explores how text-generation models work and lists some of the most important models and applications available today.

What are text-generation models?

Text-generation models are AI models trained on vast amounts of text. When prompted, they can generate new content by rearranging and combining phrases. Such models can only generate content based on its training data – so while 'new' in the sense of its ordering, the information is never exactly new.

Results can seem human-like and can be applied in use cases ranging from creative-writing assistance to conversational AI chatbots. However, the technology also raises concerns about potential misuse for generating misinformation, impersonating others online and even legal issues relating to copyright and privacy.

Here is the AI Business list of text-generation models and the applications built atop them.

Related:12 Language Models You Need to Know

Text-generation Models

GPT-4

Creator: OpenAI

First published: March 2023

GPT-4 is OpenAI’s flagship large language model. It can generate text from both images and text inputs.

GPT-4 was designed to replace GPT-3 and GPT-3.5, one of the models used to fine-tune ChatGPT. It powers ChatGPT Plus, OpenAI’s $20-a-month premium subscription service.

Not much is known about GPT-4's size and underlying details. OpenAI has, unlike previous GPT models, opted to keep things under wraps. However, the model reportedly is massive at 1.76 trillion parameters - far larger than previous GPT iterations.

Read the technical paper and news about GPT-4 on AI Business.

GPT-3

Creator: OpenAI

First published: June 2020

One of the most well-known text-generation models in the world is GPT-3

While now outdated compared to GPT-4, GPT-3 was routinely tweaked and fine-tuned to be applied to other use cases and applications, such as code generation via Codex.

GPT-3 powered a host of applications during its lifetime, including the text-based adventure game AI Dungeon, VR character creation experiences with Fable Studios and social venture projects from Create Labs. Despite its many deployments, however, Microsoft held an ‘exclusive license’ to GPT-3.

Related:AI Image-Generation Models and Tools: The Ultimate List

More from OpenAI on GPT-3: https://openai.com/blog/gpt-3-apps

Read the technical paper and news about GPT-3 on AI Business.

PaLM

Creator: Google

First published: April 2022

Current version: PaLM 2

Google PaLM is the search giant’s flagship large language model. PaLM 2, unveiled at I/O 2023, replaced the LaMDA large language model in powering its Bard chatbot application.

PaLM 2 supports more than 100 languages and is designed to be fine-tuned for domain-specific uses and applications. Take Sec-PaLM, which was tweaked for cybersecurity use cases, or Med-PaLM-2, which can help clinicians determine medical issues with images, like X-rays. 

Similar to OpenAI’s GPT-4, Google has kept mum on PaLM 2’s size. The initial iteration of PaLM, unveiled in April 2022, boasted 540 billion parameters and was designed to be used in conversational applications.

Read the PaLM and PaLM 2 technical papers, plus related news on AI Business.

Claude

Creator: Anthropic, an AI startup founded by former OpenAI engineers. Google and Salesforce are among its backers.

First published: March 2023

Current version: Claude 2

Claude is a chatbot application similar to ChatGPT and specifically designed to produce safer outputs.

It uses an approach called ‘constitutional AI’ to prevent it from generating potentially harmful outputs by giving the model a set of principles to follow.

Related:AI Code Generation Models: The Big List

Claude boasts a hefty context length, spanning some 100,000 tokens of text – or around 75,000 words – enabling it to ingest long documents.

Access the Claude API. Try Claude in Slack. Read more on AI Business.

Inflection-1

Creator: Inflection, a Silicon Valley-based AI startup backed by Bill Gates, former Google CEO Eric Schmidt and LinkedIn co-founder Reid Hoffman

First published: June 2023

One of the newest AI models on this list, Inflection-1 was built to power the startup’s virtual assistant application Pi.ai.

The startup used proprietary technical methods for the model to perform on par with the likes of Chinchilla from DeepMind and Google’s PaLM.

Try Inflection-1 via Pi.ai. Read the technical paper and news about Inflection-1 on AI Business.

Einstein GPT

Creator: Salesforce

First published: March 2023

Einstein GPT can generate personalized emails for salespeople and produce responses for customer service professionals to use when handling customer queries.

It was built by Salesforce, combining the company’s proprietary data with enterprise-grade technology from ChatGPT-maker OpenAI.

Read more about Einstein GPT on AI Business.

Cohere Command

Creator: Cohere

First published: June 2022

Current version: Command-nightly

Command is Cohere’s flagship AI generation model. It can generate text from customized data sources, meaning businesses can tune it for specific use cases.

Command can produce product descriptions, help draft emails, suggest example press releases, and answer multiple questions about documents.

It is one of the few AI models on this list to routinely evolve — Cohere releases a new version of the Command-nightly model every week.

Bloom

Creator: BigScience, a Hugging Face-led research workshop encompassing more than 1,000 researchers from 60 countries

First published: July 2022

Bloom is a 176 billion-parameter, multilingual AI model that is both open source and general purpose. Bloom can generate text in 46 natural languages and is competent in 13 programming languages. 

Access the Bloom API. Read the technical paper and related news on AI Business.

LLaMA

Creators: Meta, FAIR

First published: February 2023

Current version: Llama 2

LlaMA - or Large Language Model Meta AI - is an open source language model from Meta. It’s designed to be built upon and has formed the basis for a plethora of open source models including Alpaca and Gorilla, although LLaMA is open only for research use.

LLaMA models vary from seven billion parameters up to 65 billion parameters with the smaller models favored by researchers due to less intensive compute needed to run it.

Notably, LLaMA’s 13-billion parameter model outperforms OpenAI’s GPT-3, which has 175 billion parameters, “on most benchmarks,” according to Meta’s paper.

Meta is making the latest version, Llama 2, free for research and commercial uses. It comes in three sizes: seven billion, 13 billion and 70 billion parameters.

Read the LLaMA and Llama 2 technical papers. Read news about LLaMA and Llama 2 on AI Business.

Jurassic

Creator: AI21 Labs, a Tel Aviv-based OpenAI rival

First published: August 2021

Current version: Jurassic-2

Jurassic offers customizable language models designed to power natural language use cases. The current version, J2, is accessible via AI21 Studio, the company’s NLP-as-a-service platform.

J2 comes in a family of three models – Large, Grande and Jumbo – as well as instruction-tuned language versions of both Jumbo and Grande. 

Read the technical paper and news about Jurassic-2 on AI Business.

Luminous

Creator: Aleph Alpha is a German AI startup

First published: April 2022

Luminous can produce text from natural language prompts. 

There are four sizes to Luminous: Luminous-base is the smallest, designed for low-cost applications; Luminous-extended is the middle model designed for use cases requiring speed and can also accept images as inputs; Luminous-supreme can only handle text and is more tailored for creative writing; and Luminous-supreme-control is the largest and most expensive to run version, optimized for question-and-answering and natural language inferencing.

Try Luminous. Access the Luminous Python API and read more about Luminous.

GPT-NeoX

Creator: EleutherAI, a nonprofit AI research lab studying large language models

First published: February 2022

GPT-NeoX is a 20-billion autoregressive parameter language model trained on The Pile, an open source language modeling dataset.

GPT-NeoX is a powerful few-shot reasoner, meaning it learns well from a small number of examples.

Despite its small size, GPT-NeoX can perform a range of language-understanding, mathematics and knowledge-based tasks.

Read the technical paper. Access the training and evaluation code and model weights.

AI text-generation applications

ChatGPT

The most famous entry on this list is OpenAI’s ChatGPT, the application that started the generative AI wave and took it into hyperdrive. It turned the heads of consumers and workers across the world, opening doors to new ways of working and creating content.

ChatGPT can summarize text, answer questions and help with typing tasks. It can also help with coding questions and even execute coding tasks thanks to a Code Interpreter plugin accessible via ChatGPT Plus, its premium $20-a-month subscription service.

Since its November 2022 launch, ChatGPT has seen integrations into Microsoft and Salesforce services, among others, and has amassed 8.6 billion total visits from December 2022 to June 2023.

Read more about ChatGPT on AI Business.

Bard

Bard is Google’s answer to ChatGPT. Created to compete with Microsoft and OpenAI, Bard is powered by PaLM 2, Google’s flagship large language model.

Bard can generate text and code as well as summarize documents. It also boasts code-generation capabilities, and users can export generated Python code to Replit.

The chatbot supports more than 40 languages. 

Try Bard. Read more about Bard on AI Business.

HuggingChat

HuggingChat was created by Hugging Face, a machine learning tools developer and home to one of the internet’s largest libraries of natural language processing AI models.

First published in April 2023, HuggingChat is the open source alternative to OpenAI’s ChatGPT. Hugging Face said it built HuggingChat to keep AI text generation platforms “‘transparent” and “‘accountable.”

HuggingChat is powered by a modified version of Meta’s LLaMA model.

Alongside text, HuggingChat can generate code, although Hugging Face has released its own dedicated coding model, StarCoder.

Try HuggingChat. Read more about HuggingChat on AI Business.

Cohere

Cohere created Cohere Command. The platform gives users access to a wide range of text generation tools, including interactive chat features, product description creation and text for search and content moderation.

Cohere offers business plans, with the likes of Spotify, Oracle and Glean among its users.

Try Cohere.

GooseAI

GooseAI bills itself as a more affordable alternative to other AI generation outlets, claiming to offer a fully managed NLP-as-a-Service delivered via API at 30% of the cost of traditional providers.

GooseAI users have access to a host of AI models powering its tools, called 'the gaggle' in reference to a group of geese. These include GPT-Neo 1.3B, GPT-J 6B, Fairseq 6B and GPT-NeoX.

Try GooseAI.

Flowrite

Flowrite offers text-generation tools for business use cases, with offerings specifically tailored to salespeople and HR staff, among others.

The likes of Stripe, Snapchat, Dropbox and HBO are among its users. Its tools can be integrated into Outlook, Gmail and LinkedIn.

Try Flowrite.

InferKit

InferKit offers a web interface and API for AI–based text generators. It can be used by either developers or non-technical staff members.

InferKit is subscription-based, with tiers limiting the amount of generated characters.

Try InferKit.

Writesonic

Writesonic focuses on generating SEO-optimized marketing for blogs, online ads or sales emails.

The likes of TikTok, Schneider Electric and Amazon are among Writesonic’s users.

Try Writesonic.

Copysmith

Copysmith offers an AI content generator to support marketing and e-commerce teams. Users can import things like product catalogs to improve their workflow.

Target, Ocado and Marshalls are listed as Copysmith customers. Copysmith also has integrations for Amazon, Shopify and Google Ads.

Try Copysmith.

Jasper

Jasper not only provides audio generation tools, it also offers natural language generation options.

Jasper offers a generative API so users can bring the tools into their own platforms. It also has browser extensions and a business-oriented platform to help generate content.

Airbnb, Volvo, Hubspot and HarperCollins are among Jasper’s customers. 

Try Jasper.

Grammarly 

Grammarly is an AI tool that checks grammar, reviews spelling and punctuation and also offers tips regarding content clarity. It can generate text and detects plagiarism as well.

Clients such as Zoom and Caterpillar can customize Grammarly to check for a specific style and tone in their content.

Grammarly can be accessed via a web browser,  as an extension on Chrome, or integrated into apps like Microsoft Word. 

Try Grammarly.

Read more about:

ChatGPT / Generative AI

About the Author

Ben Wodecki

Jr. Editor

Ben Wodecki is the Jr. Editor of AI Business, covering a wide range of AI content. Ben joined the team in March 2021 as assistant editor and was promoted to Jr. Editor. He has written for The New Statesman, Intellectual Property Magazine, and The Telegraph India, among others. He holds an MSc in Digital Journalism from Middlesex University.

Keep up with the ever-evolving AI landscape
Unlock exclusive AI content by subscribing to our newsletter!!

You May Also Like