
After the major announcements from OpenAI, Google and Microsoft last month, it's time to revisit our AI Chatbot Battle and see how the major chatbots compare as of June 2024. As usual, we'll start with the free chatbots available to everyone, and then examine the paid versions.
Free AI Chatbots

The most significant change since our last comparison is that that the free version of ChatGPT now uses the latest GPT-4o LLM (Large Language Model), replacing the outdated GPT-3.5. GPT-4o is the most powerful text AI model available today, and is currently at the top of the LLM leaderboard. In addition, OpenAI has opened up many of ChatGPT's paid features to all users, including file upload, image analysis, data analysis, Internet access, and using Custom GPTs. However, free users are limited to around 10 prompts every 3 hours (this limit changes based on server load), and then ChatGPT reverts to GPT-3.5.
Microsoft Copilot still uses the previous version of GPT-4 called GPT-4 Turbo, but will soon upgrade to GPT-4o as well. They've added the ability to upload files, but the size is quite limited - only 1 MB vs. 512 MB for ChatGPT. Copilot supports image generation with DALL-E 3 in its free version, which is a major advantage, and is connected to the Internet through Bing. It also lets users run (but not create) "Copilot GPTs", which are Microsoft's version of Custom GPTs. However, there is no "GPT Store", so users are limited to one of the 5 Copilot GPTs offered by Microsoft, or Copilot GPTs offered by developers (I haven't seen any...). Each Copilot session is limited to 30 messages, but then you can start a new session immediately and continue chatting.
While the paid version of Gemini has been updated to Gemini 1.5 Pro, Gemini's free chatbot is still based on Gemini 1.0 Pro, which is weaker than GPT-4. The free version also lacks support for file upload and Custom GPTs, and the image generation feature (using Google Imagen) is quite limited: It works only in English, and refuses to generate images of people. On the positive side, Internet connectivity is powered by Google (vs. Bing for ChatGPT and Copilot), it doesn't have any usage limits, and works very fast.
In March, Claude updated the LLM which powers its free version to Claude 3 Sonnet, added the ability to analyze images, and increased the context window size to 200K, enabling users to process large amounts of text. Claude supports uploading 5 files of up to 10 MB each in a single prompt - for a long time it was the only free AI chatbot supporting file uploads, but now this feature is also available in the free versions of Copilot and ChatGPT. On the down side, Claude lacks all the other advanced features of AI chatbots - web browsing, image generation, data analysis and Custom GPTs.
The new kid on the block is Meta AI, based on Meta's open-source LLM LLama 3. It is very thin on features, supporting only image generation and Internet connectivity, but Meta is leveraging the huge user base of Facebook, Instagram and WhatsApp by integrating directly into these platforms (although there has been some user backlash regarding this integration...). Meta AI is currently available in the US and 15 other countries.
Bottom line for the free versions: ChatGPT currently has the best LLM engine and the most features, so it should be your first choice, but when you run out of GPT-4o prompts, move to Copilot. Gemini and Claude can be good alternatives for some tasks - a great technique which I teach in my prompt engineering workshops is to try the same prompt with different chatbots and compare the results.
Paid AI Chatbots
The major AI chatbot vendors have introduced new tiers to their paid plans, so I had to split my paid AI chatbots table into two parts: The first includes ChatGPT an Claude, and the second includes Gemini and Copilot. Meta AI currently doesn't have a paid plan, but it's rumored to be working on a paid version as well.

The number one question I get since the introduction of the new free version of ChatGPT is "Why pay for ChatGPT when all the features are now free?". As a paying user of ChatGPT, here are a few reasons I use to justify my own payment:
The paid version of ChatGPT supports image generation using DALL-E 3. With the chat interface you can tweak images using natural language, without rewriting the whole prompt, and they recently added a built-in full-screen AI image editor which lets you be even more creative.
The free version of ChatGPT lets you use Custom GPTs, but with the paid version you can also create them. I've already built a lot of Custom GPTs which I use daily for both work and fun.
The message cap on the paid version of ChatGPT is "up to 5x higher" than the free version - which currently means around 50 messages every 3 hours, more than enough for my own needs.
Paid users get early access to new features of ChatGPT. Right now this includes the ChatGPT Mac Desktop app and the new Data Analysis features which let you access files on Google Drive and OneDrive, interact with live spreadsheets in full screen mode, and customize your graphs. Next in line is the low-delay conversational voice mode that were demonstrated by OpenAI last month.
In addition to ChatGPT Plus which is targeted at individual users, there's a Team version with even higher capacity, that let's you share Custom GPTs among team members, and an Enterprise version that offers a larger context window, improved security and better support.
Anthropic upgraded the LLM in the Claude paid version to Claude 3 Opus, which topped the LLM charts for a while (until GPT-4o was introduced). They also added a Team version with admin tools, but as you can clearly see from the table above, both versions are lacking major features such as Internet connectivity, image generation and voice.

At Google's I/O event, the company announced that it is upgrading the paid tier of Gemini with its newest LLM - Gemini 1.5 Pro. This LLM, which has been in Beta for the past few months, boasts an amazing 1M token context window, which means you can process 1,500 pages of text in a single prompt. I've been testing Gemini 1.5 Pro Beta in Google AI Studio, and it's really good at "needle in a haystack" kind of tasks on very large documents. It also supports video analysis (you can see a demo I did with this feature here), but unfortunately video support has not been included in the latest upgrade of the Gemini paid chatbots.
It's quite confusing to call your LLM and your chatbot by the same name, but Google's pricing and positioning for the Gemini paid chatbots is even more confusing... For individuals, the Gemini Advanced chatbot is a feature of the AI Premium tier bundle of Google One, which includes Google Workspace and 2 TB of storage. For businesses, Gemini Business and Gemini Enterprise are add-ons to existing Google Workspace plans (Business Starter, Business Standard, Business Plus and Enterprise). In all of these plans Gemini is integrated into Workspace, enabling you to ask questions about your emails, documents, slides, and Google drive files. From my own experience using Gemini for Workspace integration in Google Workspace Labs, it's not working very well.
Microsoft is offering Copilot Pro for individuals, and Copilot for 365 for companies. In both cases file support is upgraded by 10x compared to the free version (10 MB instead of 1 MB), but this is still way behind ChatGPT's 512 MB files and Gemini's 2 GB files. Both paid versions are integrated with the Microsoft 365 productivity suite, so you can use Copilot to create Word documents, Powerpoint presentations, Excel formulas and emails in Outlook. Copilot for Microsoft 365 adds a very powerful feature called "Microsoft Graph Grounding", which enables you it to perform tasks such as "create a report of Q4 sales" by gathering data from One Drive files, 365 documents, Outlook emails and Teams chats. I haven't tried it myself so I can't testify whether this actually works...
Bottom line for the paid versions: If you plan on using Data Analysis and/or Custom GPTs, ChatGPT paid versions should be your first choice. Then, deciding between Copilot and Gemini is easy: You'll probably use Copilot if you're a Microsoft 365 user, and Gemini if you're a Google Workspace user. I will caution again that both the Workspace and the Microsoft 365 integrations are still under development, and features don't always work as demoed.
That's it for now - you're welcome to follow me on LinkedIn for the latest updates, and if you've found this comprehensive AI chatbot comparison helpful, feel free to share it with your network and spread the AI love!
Commentaires