Chat GPT: Plus? No. Team? Yes. 2 reasons


Chat GPT: Plus sieht alles, Team ist besser
Chat GPT: Plus sees everything, Team is better

Comparing Chat GPT Plus vs. Chat GPT Team, Team wins decisively as the sensible baseline pricing model – because of 2 reasons. Not sure which Chat GPT Version to pick? Read further.

In a hurry? Watch the video

Chat GPT Plus vs Team: 1 minute explainer video. Don’t be to critical while watching, it’s a super-quick test I made with InVideo Studio and an AI voice.
Table of content

What is Chat GPT?

Chat GPT is an AI-powered chatbot from Open AI, built on the foundation of various large language models (LLMs) such as GPT 3.5, GPT-4, and in the future GPT 5 or whatever. At its core, it generates (mostly) text, based on the user’s input, known as a “prompt.” The responses are crafted from training data, typically sourced from the internet, which may vary in terms of accuracy und uptodateness. In the learning process of most LLMs, humans are involved, to sort out issues, but that creates new issues, like human bias, or, as the finetuning guys aren’t paid well (think of way below 30 USD/h), those “teachers” are just normal people and not experts.

Many people interact with ChatGPT through the website (chatgpt.com), but it also offers an API, allowing integration into other chatbots, apps, and more. Microsoft’s Bing Copilot, for instance, used Chat GPT under the hood, but not with standard settings. First, it runs on Microsoft’s own servers (MS pours truckloads of money in AI server farms & cloud), and secondly, with customized settings. As a result, the capabilities and outputs can differ significantly.

Although Chat GPT is the best known offering, thanks to brillant marketing (guess what, Open AI employs wordclass advertising pros), it isn’t alone in the world of Generative AI. There are numerous LLMs and AI offerings available, some even open-source (partially, like Mistral from France) or tailored for security-critical applications (such as Anthropic Claude, which has GOVT/MIL solutions).

Training and Running an AI chatbot is expensive on an eye-watering level, which means that professional services often come with a price tag. Setting up a LLM from scratch is rumoured to cost 1 bn USD, and finetuning an existing LLM, e.g., in case you are interested to use an open source LLM to create your own corporate LLM, might cost some millions of USD, alone in server costs (= those fancy server farms full of NVIDIA hardware).

Unfortunately, and unsurprisingly, the industry has gravitated towards the subscription model, saddling businesses with monthly fixed costs. You know that from Microsoft and Adobe: Pay, peasant, no matter of you use the services or not.

Let’s take a closer look at the subscription plans available for ChatGPT.

Chat GPT offers four pricing models

ChatGPT: Cost comparison

The cost for Chat GPT is, for Plus and Team, a monthly price per user (or “seat”). The Enterprise plan has no official price list, thus you have to negotiate on your own (good luck with this one). The fourth cost structure is the “Pro” level and shows the real, unblended cost: the price for API calls, respectively the price per token. Every token represents a share of computing power used, which consists of the (steep) prices for building server farms, training LLMs, energy, and, erm, profit.

As a heavy user of the the ChatGPT Team plan you will likely make a good deal, but if you’re going Enterprise and your organization simply doesn’t use it, you’re wasting money. Thus you better do some internal research what tasks your staff really could and will “outsource” to an Gen AI assistant – or probably a set of specialized AI tools and No-Code-Apps will be a better fit. Always remember that ChatGPT only offers its own LLM, while independent app or solution providers are free to use whatever LLM is best for the task.

By the way, the API is getting more important: The new features of GPT-4o (Audio / “Call Center Chatbot”, Video) will be released a) first for API customers and b) work best inside an app.

Chat GPT StandardChat GPT PlusChat GPT TeamChat GPT EnterpriseOpen AI API Calls1
Free20 USD
/ month
30 USD
/ month
ca. 60 USD
/ month
ca. 0.5 – 120 USD
/ 1 Million Tokens

1. Chat GPT Free (the No-cost we all know)

The standard free version relies on the GPT-3.5 model. Often glitchy and yielding strange results, it’s rarely suited for serious professional use in Non-English languages (e.g., German).

Now, with GPT-4o (the “o” is for Omni, meaning Jack of all AI trades), the free version got a remarkable upgrade, as the free plan (partly) uses GPT-4o.

What has been improved?

  • ChatGPT-4o in the Free version has (partly) access to the internet: It uses the Bing search, but translates e.g., a German input to English first. This way, the results are more factual, but overall not reliable. The overall output resembles Microsoft’s Bing Copilot. The downside with Bing is that the search results
  • Internally, the web access is triggered by (automatic) pre-prompting: When you ask about people, ChatGPT will add the $currentyear to your input, which triggers a web search. You can do that on your own if you add the current year to your prompt. This way, the output will contain actual links to websites – which, of course, can be bollocks as usual. This means: You never know if a result is usable or not.

2. What is Chat GPT Plus? The Basic Subscription Model

Chat GPT Plus runs $20 a month, granting access to superior GPT-4 (and GPT-4o) and the ability to create custom templates (GPTs). Moreover, it browses the web and generates images (DALL-E)
You want to try Chat GPT 4 before you buy?
Using Bing Copilot lets you sample Chat GPT 4, particularly using the Copilot apps. Bear in mind, Microsoft hosts Copilot on Azure servers and employs unique foundation prompting. Out of the box, Copilot heavily depends on search results and footnotes, causing noticeably different generation outcomes.
Moreover, Microsoft struck an exclusivity deal with Mistral AI for customers to access “Mistral Large” (pronounce it French). Mistral’s large language models produce European-feeling results, whereas Open AI LLMs tend to sound like translations from English. Fun thing, this text was translated to English from German with a Mistral LLM, and it gives you that special “translated” feel – of course on purpose to create that special Euro vibe.

3. What is Chat GPT Team? The Mid-Range Enterprise Solution

Chat GPT Team resembles Chat GPT Plus but charges at least $60 monthly with a minimum of two user licenses ($30 each). Total expense reaches $720 yearly or approximately €670 (as of Feb 2024). Annual advance payment lowers the rate to $25 per user, totaling at least $600 initially. Despite the sum, it remains the sole viable starting point for business use
Otherwise, take advantage of the complete Chat GPT Enterprise edition – more info on that shortly after the break).

Here are two reasons not to pick Chat GPT Plus

1. Chat GPT Team avoids using your data as training data

Every piece of information you submit or request through Chat GPT Plus (or Free) gets recorded as training data again (yes, you can exclude your input from training data use, but then your history is deleted). This is absolutely unacceptable and the reddest of all red flags.

Why is that problematic? If proprietary company information serves as source material, it ends up stored within the vast database of the big language model. From there, competing firms could incidentally or deliberately extract your secrets. This discussion goes beyond mere legalities or General Data Protection Regulation (GDPR, or DS-GVO as we Germans know it, “Datenschutzgrundverordnung”); actual complications arise here.

Imagine obtaining detailed tech specs for a revolutionary invention from your Technical Development. You intend to condense them into bullet points and develop sales pitches for a meeting. Unfortunately, entering raw data feeds the language model classified intel. Awkward situations occur when rivals research comparable fields, and engineers query Chat GPT for innovation. Surprise: The system churns out your secret project. Luckily, with Chat GPT Team, this risk supposedly vanishes. Many corporations host their Chat GPT instances on local servers to safeguard intellectual property.

2. Craft team-oriented GPTs with Chat GPT Team

Utilizing AI tools within businesses generally focuses on streamlining repetitive tasks, reminiscent of the industrial era introducing machinery. GPTs represent the mechanical limbs of generative AI. They help build templates that benefit the whole team, thereby improving outputs and conserving time. Not everyone possesses the talent, interest, or patience to master prompt engineering. Thus, having only one competent teammate (or outsourcing to freelancers) grants all teammates access to helpful mini-apps.

To disseminate these GPTs among your crew, you’ll require Chat GPT Team, as Plus simply won’t cut it.

One remaining concern centers on finding the tangibly beneficial use case for your firm. Identify this first, possibly via employee polls (Do you use Chat GPT? Any frustrations? Time-consuming activities?) or AI workshops. Then pinpoint the perfect tools for your needs.

4. What is Chat GPT Enterprise? Expensive: Over U$D 108,000 or €100,000 p.a.

No official price list exists for the premium Chat GPT Enterprise. One must contact sales for specifics. Allegedly, the cost starts at $60 per month with a minimum commitment of 150 single-user licenses, amounting to $9,000 monthly or $108,000 yearly.

Primary Advantage: Operate the language model on your private cloud, engage with APIs, and integrate it with Single Sign-On into your corporate IT security policy.

Alternatives to Chat GPT? Check out Office Copilot, Hugging Chat, iStock, Bria.

  • For regular office duties, Microsoft Office 365 Copilot might appeal given its similar pricing, and featuring neat features such as converting Word docs into PowerPoints. Another bonus: thread summary within the inbox. Interestingly, Microsoft hosts the Open-AI language model on their own servers, with user data protected against repurposing for training. Sounds good, right?
  • Image manipulations fare safely with Bria.ai, considering licensed training data. But stay vigilant, as seemingly “editorial-only” stock images (like those from Unsplash) can be used there.
  • Exploring AI-generated imagery, Adobe Firefly and iStock stand out, both with subscription respectively paid (credits) models. NVIDIA Picasso powers iStock creations, ensuring license compliance. Don’t count on flawless photographic quality. Bria and Firefly shine brightest in generating illustrated artworks.
  • Open-source language models offer another exciting opportunity. Build and host your GPT, protecting data privacy. Test freely with Hugging Chat Assistants by Hugging Face, resembling Chat GPT’s Custom GPTs. Need inspiration? Browse their site for prefab chatbots.
  • To create “apps” or to automate processes, you can check out Microsoft’s “Power Automate.” It offers low-code / no-code ready-to-go mini-apps like “Analyze sentiment in emails with AI Builder and send results to Teams.” This automated workflow a) leverages Microsoft Dataverse, b) reads emails in Outlook, c) analyzes the content (sentiment analysis), and d) posts the results in Teams. This is particularly interesting for customer service, allowing you to quickly route complaints to the right people. You can also link Power BI and Visio, but for that, you’ll need skilled IT consultants.
  • To transform prompts into apps, essentially shoehorning the prompt as an API call into Python or something similar, Anthropic is pretty much the bee’s knees with Claude. However, that’s when things start to get nerdy.
    Pro tip: Anthropic’s “workbench” has a prompt engineering tool that is quite interesting.
  • Long-term goal: leverage open-source speech models, host independently, and personally train with curated data. Reserve app use solely for employees, excluding internal affairs and global politics. For public website chatbots, consider a second system. After all, chatbots excel at knowing your goods thoroughly, just refrain from political discourse.
  • Address challenges inevitably, say cleaning raw data before uploading personal details to prevent excessive exposure.
  • Curious about fine-tuning custom chatbots? Review Eden AI’s article covering “trainable” systems like Eden AI, Chainlit, and AskYoda.

Summary

Among the pricing models, only Chat GPT Team or Chat GPT Enterprise ensure relatively clean utilization for demanding tasks involving sensitive data, unlike Chat GPT Plus.

Note: Prices may vary. Privacy policies and licensing terms apply.

Stefan Golling

About the Author

Stefan Golling, Cologne, Germany. Worked since 1998 as a Copywriter and Creative Director in (Network) Agencies and freelances since 2011 as German Freelance Copywriter, Marketing Freelancer, Creative Consultant etc., e.g., in international projects.

Ask for AI consulting in the field of marketing


  1. ↩︎
⌀ 5. Stimmen: 4
Zu Hause » Mag » Chat GPT: Plus? No. Team? Yes. 2 reasons

Erstellt am:

Zuletzt aktualisiert: