The Ultimate LLM Comparison: Which Is Right For You?

Discover the Evolution of AI
Explore the transformative journey of artificial intelligence from its beginnings to future innovations. The chatbot competition is an AI-fueled arms race — fortunately, arms used to create and educate.

Will this contest lead to more advanced, equitable technologies, or will unforeseen obstacles arise? How ‘human’ can AI truly become? Will AI tools become an indispensable, everyday part of society?

Today, at least one essential question will hopefully be answered: which generative AI chatbot is best for your needs? The following six AI tools are today’s main players for research and generative text — and at least one of them may prove to be an indispensable tool in your arsenal.

TL;DR What’s the best functionality of each chatbot?

• Google Gemini: creative inspiration
• OpenAI ChatGPT: versatility
• Anthropic Claude 2.1: extensive content analysis
• Microsoft Copilot: Microsoft integration
• xAI Grok: witty & real-time information
• Perplexity: web information retrieval

By pinpointing your core requirements, you can effortlessly select the chatbot best aligned with your objectives.

Gemini by Google

Released in March 2023, Google Gemini is designed to pique curiosity, augment imagination, and inspire fresh ideas.

Available in over 230 countries and regions in more than 40 languages, it can translate languages, effortlessly handle idioms and other nuanced text, and even offer personalization, with customizable responses in 5 styles: simple, long, short, professional, casual. With the ability to listen and speak, Gemini is perfect for audio learners, pronunciation help, and hearing poems or other literature.

Gemini now integrates with Google apps, providing real-time information from Maps, YouTube, Hotels, and Flights. Plus, its adoption of implicit code execution in June 2023 made Gemini a truly helpful tool for solving complex, computation-based word and math problems.

The tool also excels at over 20 programming languages, including specialized ones like Prolog, Fortran, and Verilog and can also share conversations, pin responses, and analyze images using Google Lens.

How does Google Gemini work?

Gemini initially used Google’s next-gen language model, Pathways Language Model 2 (PaLM 2).

PaLM 2 is based on Google’s Transformer — a type of deep-learning neural network architecture commonly used in natural language processing (NLP) — which provided the framework for GPT-3 (Generative Pre-trained Transformer 3).

Formerly recognized as Bard and recently rebranded as Gemini, this suite of multimodal large language models (LLMs) has been crafted by Google DeepMind, positioning it as one of the most sophisticated chatbot technologies currently accessible.

How are people using Google Gemini?

With its plethora of capabilities and data from Google, Google Gemini is useful for a variety of tasks, including (but certainly not limited to):

Imagery

Gemini's Google Lens-powered image capabilities allow it to transcribe letters or other visual materials.

Research

Able to explain difficult concepts (such as quantum mechanics) in simple terms, Gemini is an adept study or productivity companion.

Content

The chatbot can also provide content outlines for blog posts and other content.

IT help

IT-wise, after uploading an image of an error or confusing message, Gemini supplies explanations, instructions, and fixes.

Travel & recipe planning

An adept travel planner, Gemini suggests flights, hotels, and activities based on budget and user preferences. It can find online deals. Appetizingly, the tool recommends recipes based on an image of a user’s ingredients.

Pros of Google Gemini

• Supplements Google search — harnessing Google’s relevancy, scope, & location-sensitivity
• API features for incorporation with various applications, including translators, content generators, conversational chatbots, Q&A platforms, messaging platforms & websites
• Can speak and listen
• Generates multiple answers in ‘View Other Drafts’ option

Cons of Google Gemini

• Like other generative AI systems, deals with misleading information and inaccurate data
  • During a February 2023 demo, it gave a wrong answer about the James Webb Space Telescope — losing Google more than $100 billion in value the next day
• May be prone to repetitive or bland responses compared to other chatbots

ChatGPT by OpenAI

Igniting the chatbot revolution circa November 2022, OpenAI’s ChatGPT stands out with its versatile applications, from generating creative content to writing code to solving complex problems​​​​​​.

Designed to mimic human cognition, it's known for its ability to engage in multifaceted, fluent conversations for a wide range of purposes, such as finding information, brainstorming ideas, and exploring creative writing.

It can also be paired with Assistants API to broaden its use cases even further. Plus, OpenAI also opened its GPT store in January 2024 — allowing anyone to create and share customized versions of ChatGPT with specified training and instructions.

How does ChatGPT work?

OpenAI’s GPT models (including GPT-3.5, GPT-4, & GPT-4 Turbo) were trained to peruse documents and then predict the following word in a sentence — drawing on mostly publicly available data that covers reasoning, contradictions, consistencies, ideologies, mathematical solutions, and fallacies.

Human reinforcement then adds guardrails and shapes the answers to align with human cognition.

What’s the difference between GPT-3.5, GPT-4, & GPT-4 Turbo?

Faster and more accurate than GPT-3.5, GPT-4 is 82% less likely to respond to disallowed content and 40% likelier to generate factual responses. It also tests significantly better in a variety of areas, including the LSAT — with GPT-4 placing within the top 10% in the bar exam and GPT-3.5 in the bottom 10%.

GPT-4 is also more creative and able to handle nuanced queries and can accept and describe image inputs, even explaining why a picture (or meme) may be humorous. Plus, its “steerability” imbues it with personalities, such as a Socratic tutor that encourages critical thinking.

Compared to GPT-4 and GPT-3.5, the latest ChatGPT iteration, GPT-4 Turbo, is able to handle larger amounts of data. Its integration with DALL-E 3 offers enhanced multimodal capabilities — enabling it to process not only text but images, audio, video, and other formats.

How to access each

• GPT-3.5: ChatGPT’s free version
• GPT-4: ChatGPT Plus

GPT-4 Turbo: “Anyone with an OpenAI API account and existing GPT-4 access can use this model by passing ‘gpt-4-1106-preview’ as the model name in the Chat Completion API,” according to OpenAI

ChatGPT is an effective option for a diverse range of activities, such as:

Websites & Content

GPT can save time by creating websites, blogs, social media content, product descriptions, or other marketing material.

Business

It can also help you draft or develop business ideas and can play the role of business assistant by researching, brainstorming, or simplifying complex ideas.

Coding

GPT can help programmers by writing or debugging code in numerous programming languages, including Python, JavaScript, Java, PHP, C++, C#, Ruby.

Education

GPT can enhance education by helping students study every imaginable topic, with instruction — for example, solving math problems with step-by-step explanations.

It can also aid teachers, employers, and marketers create quizzes, worksheets, surveys, interviews, and onboarding materials.

Pros of OpenAI ChatGPT

• Can create personalized content for education, marketing, artistic purposes, curiosity, & amusement
• Multimodal GPT-4 is capable of accepting image inputs
• Multilingual in human & non-human languages — analyzing & writing code in numerous programming languages
• API support & plug-ins

Cons of OpenAI ChatGPT

• Requires paid ChatGPT Plus subscription to access GPT-4 for quicker responses, access to upgrades, & priority during busy times
• Data set is only current to April 2023
• Like others, prone to factual mistakes & reasoning fallacies
• Does not learn from its experience
• Solutions may not be perfectly reliable, e.g. code with security vulnerabilities

Claude 2.1 by Anthropic

Anthropic, founded by siblings and former OpenAI employees, Daniela and Dario Amodei, released Claude 2.1 in November 2023.

The chatbot is known for its extended context window, which enables it to process extraordinarily large amounts of text — accommodating analysis, review, or creation of entire books or codebases.

With the company striving to address ethical and social challenges, Claude 2.1 also excels at generating accurate, reliable responses. The chatbot can also generate tables, follow Markdown formatting, and is able to write code, do math, reason, and converse with creative answers.

How does Claude 2.1 work?

Claude is an encoder-decoder language model with Constitutional AI built-in at its foundation. Akin to Isaac Asimov’s Three Laws of Robotics, Constitutional AI ensures maximum helpfulness while minimizing potential harm by producing responses that avoid illegalities or dangerous advice.

Unlike GPT-4, which relies on learning via human reinforcement, Claude uses a self-fine-tuning system — meaning it does not depend on human feedback for training.

How are people using Claude 2.1?

Valuable for an array of tasks, Claude 2.1 can help you with:

Research & Learning

Its ability to review or create books, lengthy research materials, technical documents, or codesets and offer comprehensive summaries, background information, and relevant statistics make it great for study help, book reviews, and legal or financial analysis.

Writing & Marketing

Claude is well adept at writing — and can even help write fiction novels. Also a great marketing tool, it’s able to create personalized messages, identify relevant keywords, and optimize content.

Business

Claude can quickly and accurately extract relevant information from business emails and documents as well as categorize and summarize survey responses.

Customer Support

With enhanced comprehension, longer memory, and reduced hallucination rates, the chatbot can act as an always-on virtual sales representative that boosts customer satisfaction with speedy, friendly resolutions to service requests.

Coding

Able to interpret and generate code in popular programming languages while maintaining context throughout the code, Claude 2.1 is an effective tool for beginner and intermediate-level coding needs.

Pros of Anthropic Claude 2.1

• Uniquely huge context capacity, capable of handling extraordinarily large prompts
• Designed to produce bias-free outputs via Constitutional AI, enhancing ethics
• Integrates with various apps via API support & Claude App for Slack
• Can self-learn & improve without human feedback

Cons of Anthropic Claude 2.1

• Like other AI chatbots, occasional mistakes, “hallucinations,” fallacies, & biases
• Only available in the US & UK (without a VPN)
• Unlike other chatbots, not connected to the internet

Copilot by Microsoft

First released in February 2023 as Bing Chat, Microsoft Copilot is an innovative AI chatbot that can revolutionize the way you work.

Seamlessly integrated into Microsoft 365 apps, Copilot Pro provides real-time, intelligent assistance that enhances your creativity and productivity in a myriad of ways (see below).

As of January 2024, the chatbot now gives you free access to GPT-4 Turbo & DALL-E 3 in both its Pro and free versions — with priority to Pro subscribers. GPT-4 Turbo offers expanded query sizes as well as more meaningful responses, while DALL-3 integration enables higher-quality image generation.

Copilot now also has dedicated iOS and Android apps, so you can easily access the tool even when you’re away from your computer.

How does Microsoft Copilot work?

Microsoft Copilot is powered by a combination of LLMs that are deployed and operated within the Microsoft Cloud and are not trained on organizational data — ensuring privacy and security.

In addition to these Microsoft Cloud LLMs, Copilot is powered by the same technology behind ChatGPT, running on GPT-4 Turbo and DALL-E 3 as of January 2024.

How are people using Microsoft Copilot?

Copilot is beneficial for a wide spectrum of use cases, including:

Business

Copilot’s integration into Microsoft 365 apps make it a powerful tool for streamlining work processes and boosting productivity by:

• Word: creating, summarizing, comprehending, & refining documents
• Excel: generating formula suggestions, showing insights, & highlighting interesting portions of data
• PowerPoint: drafting presentation, summarizing key points, or restructuring slides — all from a text prompt
• Outlook: transforming long email conversations into short summaries & assisting with grammar, style, and content.
• OneNote: organizing, summarizing, rewriting, & searching through your lists for you
• Teams: summarizing key points, extracting essential information, & identifying action items

A few of the many areas in which Copilot can help streamline operations include:

• Customer service: powering always-available AI chatbot support agents
• Risk assessment & management: analyzing large datasets & predicting potential issues
• Supply chain management: predicting disruptions across suppliers, weather,& geographies
• Healthcare: aiding in patient data management, medical research, & appointment scheduling

Research

Copilot’s multi-format capability enables it to create comparison tables — even researching the specs of multiple products and showing them side-by-side.

Planning

Bing Chat can also serve as a daily planner. Whether it’s a meal, inexpensive date night, movie selection, or a grand party, the chatbot can help determine venues, decorations, cuisine, activities, and more.

Coding

Microsoft’s Github Copilot revolutionizes the way developers work by providing instant suggestions that enhance code quality and consistency, auto-completing code, and catching errors before they become issues.

Pros of Microsoft Copilot

• Seamless integration across Microsoft 365 apps
• Real-time suggestions enhance productivity
• Supports numerous plug-ins & allows you to create plugins via Copilot Studio
• Integrated security and compliance, inheriting your company’s protocols

Cons of Microsoft Copilot

• Inadequate configurations may result in unintended exposure of organizational data, causing security vulnerabilities
• Like other AI chatbots, occasional mistakes, “hallucinations,” fallacies, & biases

Grok by xAI

Co-founded by Elon Musk, xAI’s Grok, was released in November 2023. This newer chatbot offers uncensored, humorous responses with an unconventional personality that’s inspired by Ford Prefect, the whimsical character from the Hitchhiker’s Guide to the Galaxy.

By being able to respond with witty ideas, Grok has the potential to foster more engaging interactions — which some consider a reprieve from other chatbots’ political correctness.

Grok’s access to X (formerly Twitter) content enables it to offer real-time information on global events. It’s also planning to offer multimodal capabilities, including audio and vision — which will make it a truly versatile AI tool.

How does Grok work?

Grok is powered by Grok-1, an LLM developed by xAI that’s built on a custom training and inference stack based on Kubernetes, JAX — a machine learning framework developed by Google, and the programming language, Rust.

According to xAI’s research, when benchmarked against leading LLMs, The Grok-1 model outperformed GPT-3.5 in many areas but not GPT 4.

How are people using Grok?

Grok’s access to real-time information and its humorous responses make it advantageous for a broad scope of use cases — a few of which include:

News & trending topics

Monitor — and get notified about — live events, breaking news, and emerging trends.

Market research

Get up-to-date insights on summaries on customer sentiments and behaviors.

Crisis management

Instantly relay imperative information during a disruptive event.

Education

Engage students on a new level with humorous lesson plans and activities.

Writing

Transform your content by learning insights and techniques for integrating humor effectively, identifying your humor style, and experimenting with various humor techniques.

Customer Service

Is your company more unconventional than traditional? Infuse your support chatbot with a voice that resonates better with your brand.

Pros of Grok

• Varied modes, ‘fun mode’ for humor & snarkiness and regular mode for more serious conversations
• Real-time information via connection to the X platform
• Capable of handling "spicy" questions typically rejected by other AI chatbots
• Multitasking capabilities allow for multiple, simultaneous conversations & conversation branching allows for digging into specific areas without disrupting the main discussion
• Built-in markdown editor enables downloading, editing, and formatting responses for later use

Cons of Grok

• Limited availability to a select group of US-based X Premium+ subscribers for testing and feedback
• Smaller context window than other chatbots limits how much information it can process — meaning it may fall short in powering intricate corporate needs & high-quality outputs
• May encounter problems from being in early stage development
• Poor training from basing information on X posts, some of which include misinformation
• Inconsistent responses that vary greatly based on how questions are phrased
• Prioritizing humor threatens accuracy

Perplexity

Founded in August 2022, San Francisco-based Perplexity aims to democratize knowledge by making searching easy and intuitive.

Perplexity can summarize a web page, answer questions about the page, provide shareable links, and generate follow-up questions. In addition to its conversational interface, its contextual awareness offers in-depth answers with added perspective, suggestions, and citations.

Perplexity also offers up-to-date information and lets you fine-tune your search by narrowing down sources through its Focus feature. Plus, it has a Chrome extension, so you can instantly access the chatbot without having to navigate to the website.

How does Perplexity AI work?

Perplexity’s basic search function utilizes GPT-3 for all search queries and text generation.

Its Pro version uses GPT-4 for more interactive, personalized experiences — asking clarifying questions, engaging in back-and-forth conversations called ‘Threads’, offering multiple searches, and summarizing results.

Perplexity also offers a Microsoft Copilot feature, which utilizes GPT-4 and incorporates WolframAlpha data, a computational knowledge engine that accesses a wide range of curated data. Access to Copilot is far more accessible with Perplexity Pro than the basic version.

How are people using Perplexity?

Because Perplexity vastly scours the web to retrieve real-time information, it’s beneficial for a wide range of tasks, including:

Research & Education

Perplexity offers customized research assistance for students and scientists writing, analyzing, or summarizing research.

Education

Students can use it to digest difficult topics, solve mathematical or other assignments, and write essays.

Market Research

Perplexity can summarize industry trends, global news, and pop-culture happenings.

Pros of Perplexity

• Contextual, conversational, & personalized answers offer deeper comprehension
• Combines prominent LLM & search functions, marrying AI technology from OpenAI and Microsoft
• Reviews conversation history to learn your preferences & personality
• AI Profile feature remembers preferences & tailors the experience
• Ability to generate or process text & code without internet access

Cons of Perplexity

• Pro plan required for advanced features, such as unlimited file uploads, dedicated support, & Copilot & GPT-4 use
• Like other tools, prone to hallucinations & falsities