I spent $200/mo on ChatGPT Pro so you don’t have to. It wasn’t worth it | usagoldmines.com

What you get with ChatGPT Pro

What exactly do you get with ChatGPT Pro that you can’t get in the free version? Or even the much cheaper ChatGPT Plus plan that only costs $20 per month? The short answer is, every single feature OpenAI has to offer — and that list continues to grow quite quickly.

The longer answer is that ChatGPT Pro provides access to all OpenAI LLMs (including GPT-4o, GPT-4.5o, o3-mini, o3-mini-high, and o1) along with all the latest features like Operator and Deep Research.

On top of that, ChatGPT Pro includes full access to the company’s AI models for media generation, which includes Dall-E for image generation and Sora for video generation. Pro users get priority video generation and, in the case of Sora, access to better resolutions (up to 1080p) and extended durations (up to 20 seconds).

ChatGPT Free vs. Plus vs. Pro

If you’re still feeling a bit confused about what you get with the different ChatGPT plans, don’t worry, you aren’t alone. Here’s a quick chart I whipped up to help you visualize the differences:

Matt Smith / Foundry, made with Claude

As the chart shows, upgrading from ChatGPT Free to ChatGPT Plus unlocks access to multiple features, while upgrading from ChatGPT Plus to ChatGPT Pro only unlocks one wholly exclusive feature: Operator.

However, I want to draw your attention to the last feature listed in the comparison, which is rate limits. OpenAI regularly changes the rate limits for ChatGPT Free, Plus, and Pro, and they aren’t always documented.

ChatGPT Free

Generally, though, the rate limit for ChatGPT Free is extremely low. You can exhaust resources with just a few conversations, forcing you to wait hours before you can ask another question. ChatGPT Free’s rate limits for more advanced features (like the o3 reasoning model and Dall-E image generation) are so low that they border on useless.

ChatGPT Plus

ChatGPT Plus is much more generous. While OpenAI’s LLMs still have firm rate limits in this tier, they’re high enough that most people won’t hit them with typical usage. The rate limits for other advanced features (like Dall-E, Sora, and Deep Research) are still tight, though.

ChatGPT Pro

ChatGPT Pro’s rate limits are much higher. Certain advanced features are still kind of limited (e.g., Deep Research is capped at 120 generations per month) but high enough that most people won’t come close to hitting them. I used ChatGPT Pro heavily, racking up more than a dozen full conversations a day, and never hit any rate limits.

In other words, the rate limits on ChatGPT Pro will only become an issue if you’re incorporating it into automated workflows — in which case you’re really meant to access OpenAI’s models through its APIs.

ChatGPT Pro vs. Claude with Perplexity, Midjourney, ElevenLabs, and Kling

OpenAI is the most famous of AI companies, but it’s far from the only one. Its many competitors include DeepSeek, Anthropic, Midjourney, ElevenLabs, Meta, and Google to name a few.

Which raises the question: If ChatGPT Pro is so dang expensive, how does it compare to the competition? Well, the $200/month spent on ChatGPT Pro would actually be enough to cover the monthly plans for several different AI alternatives, spreading the cost across multiple.

For this review, I tried ChatGPT Pro against five alternative services:

• Anthropic Claude Pro (AI chatbot) — $20/month
• Perplexity Pro (web search) — $20/month
• Midjourney Standard (image generation) — $30/month
• Kling AI Pro (video generation) — $25/month
• ElevenLabs Creator (speech-to-text/text-to-speech) — $22/month
• Total cost — $117/month

Choosing a suite of competitive AI services with capabilities similar to ChatGPT Pro certainly looks appealing on paper. The total cost is nearly half the price, a steep discount compared to ChatGPT Pro.

Matt Smith / Foundry

This à la carte approach arguably provides access to a better selection of AI services, too. Coders generally prefer Claude over ChatGPT while Perplexity is better for web search than ChatGPT’s built-in search, for example. Creatives hugely prefer Midjourney’s image generation model to Dall-E, and Kling AI’s titular video models are often recommended over Sora. ElevenLabs, meanwhile, offers speech-to-text and text-to-speech capabilities that ChatGPT doesn’t even provide.

On the other hand, ChatGPT Pro offers some features that competitive services don’t emulate. Claude doesn’t offer alternatives to Deep Research or Operator. ElevenLabs’ text-to-speech and speech-to-text are useful but not the same as ChatGPT’s Advanced Voice. And while a variety of AI agents exist, no competitor offers a simple, ready-to-use AI agent like OpenAI’s Operator.

This comparison between ChatGPT Pro and a suite of alternative services shows an important point: if you’re subscribing to ChatGPT Pro, it’s likely for access to features that are exclusive to ChatGPT Pro (or rate limited in such a way that they’re only useful with ChatGPT Pro).

In that case, we need to dive into those exclusive ChatGPT Pro features and evaluate them for value. Are they worth it? Let’s find out.

ChatGPT Pro’s reasoning models

Let’s start with ChatGPT Pro’s access to OpenAI reasoning models. These models use chain-of-thought reasoning to prompt themselves as they try to work through a problem.

This helps the model whenever it’s asked to deal with something that involves logic or requires understanding the real world to reach the correct answer. Reasoning models score well across a wide range of publicly available benchmarks, but they stand out in math, coding, science, and related fields.

Matt Smith / Foundry

For example, when asked to write code for a particular feature in an app, a reasoning model will often do a better job. It can reason through potential use cases for the feature, as well as the varying approaches to implementation, as it looks for an adequate response. This is more likely to result in code that’s immediately useful without modification.

Because of this, reasoning models like OpenAI o1 tend to dominate AI benchmarks focused on math and coding.

It’s not all good news for reasoning models, though. They’re slower than conventional AI chatbots, and the more reasoning an AI does, the slower it gets. (ChatGPT o1 can leave you waiting for a minute or longer before it starts to generate a reply.) Reasoning models also tend to have more limited access to files and tools. If you want to use OpenAI o1 to analyze a PDF, for example, you can’t upload it. The best you can do is copy and paste the text into the prompt, which limits the usefulness of reasoning models depending on what you want ChatGPT to do.

Reasoning models are great if you want an AI model to help you brainstorm difficult problems or write code that’s more relevant to your software. But if you want to use ChatGPT as a writing or editing assistant, or to analyze files and information, or to answer complex questions, then reasoning models are often less useful than GPT-4.5o.

Deep Research is pretty awesome

Deep Research is a ChatGPT feature that employs an AI agent to search the web, ingest relevant information, and compile it into a lengthy summary with sources cited. That report can be used for more serious topics (like researching career opportunities in a specific industry) or for more casual pursuits (like compiling a history of cat memes).

No matter how you decide to use it, Deep Research’s reports deliver generally useful information and often pick up on details that would be hard to find when searching the web manually.

Matt Smith / Foundry

Say you want to open a restaurant. You can ask Deep Research to write a report on the restaurant industry for your specific region, and you can even prompt it to find competitors and look for gaps in the market. Deep Research can provide granular recommendations that include underserved niches you can fill, and it can make recommendations down to the neighborhood level, citing specific competitors in said neighborhood to justify its conclusions.

The reports aren’t perfect, though. Hallucinations can cause Deep Research to generate incorrect information. And because it searches the web to build its report, it’s susceptible to oft-repeated generalizations, myths, and inaccuracies. Then again, even if you tried compiling that research on your own, you’d likely run into the same issues. So, it’s up to you to decide what to trust and what to filter out.

Yes, it has its limits and flaws, but Deep Research is a great feature. The information it compiles is often more detailed than what I could hope to achieve on my own, and it tends to find deeper web sources that I might otherwise miss, skip, or overlook. Deep Research is the best feature currently bundled in ChatGPT Pro, and while competitors are rapidly trying to mimic it, none have matched it yet.

Deep Research is available with ChatGPT Plus, but Plus users are limited to 10 queries per month. ChatGPT Pro gives you 120 queries per month (as of March 2025), and that number will likely increase later.

Dall-E and Sora sadly disappoint

OpenAI’s Dall-E image generator and Sora video generator aren’t exclusive to ChatGPT Pro, but upgrading to Pro makes them more useful. It also exposes some limitations.

Dall-E

Dall-E is tightly integrated into ChatGPT. The chatbot doesn’t just generate images on request, but also whenever it thinks a visual response is warranted. But Dall-E’s quality is lacking and it hasn’t improved much over the past year, even as competitors like Midjourney continue to make progress. Most competitors also offer advanced features Dall-E lacks, such as storyboarding, detailed image editing, and concurrent jobs.

ChatGPT Free has access to Dall-E, but usage rate limits with Free and Plus plans are tight. You’ll need ChatGPT Pro if you want to generate dozens or hundreds of images in a session.

Sora

Unlike Dall-E, Sora video generation isn’t accessed through ChatGPT itself. It’s a separate tool — one that still requires a ChatGPT subscription to use. While ChatGPT Plus users get limited access to Sora, ChatGPT Pro users enjoy enhanced quality (up to 1080p) and longer videos (up to 20 seconds versus Plus’s limit of 5 seconds). Rate limits on ChatGPT Plus are tight, too, so if you’re looking to use Sora as anything more than a toy, you’ll need a ChatGPT Pro subscription.

Matt Smith / Foundry

Unfortunately, Sora disappoints. It struggles with basic tasks like stable camera pans and fails almost completely with fast action sequences. Competitors like Kling AI can produce better results while other services like Runway and Pika Labs offer specific AI models for different tasks (like character animation or special effects) with more reliable results.

Dall-E and Sora are weak spots in the ChatGPT Pro subscription. Professional artists and creators looking to make ChatGPT part of their workflow are likely to find it isn’t up to snuff.

Operator falls short of its potential

Operator is OpenAI’s advanced reasoning agent for executing simple tasks in a web browser. It’s notable for being the only feature exclusive to the ChatGPT Pro tier subscription (as of March 2025).

Like Sora, Operator is a separate tool outside of ChatGPT that still requires a ChatGPT Pro subscription to use. After entering a prompt with your requested action(s), you’ll see a real-time feed of the agent attempting to fulfill your request in a virtual machine. In the video below, you can see OpenAI’s real-world demonstration of Operator:

In theory, an AI agent like Operator could function as a virtual assistant that handles everything from online shopping to completing forms or organizing emails. Operator sometimes succeeds with these tasks.

Beyond the basics, however, Operator remains quite limited. It can’t bypass security measures like CAPTCHAs, instead prompting you to complete them. It won’t automatically enter payment details either. (The reasons for that are obvious, right? I wouldn’t trust Operator with my credit card details.) Even so, these limitations make Operator a lot less useful. As of now, you still have to help Operator over various hurdles, which defeats the point of an autonomous AI agent.

Should you pay up for ChatGPT Pro?

As this lengthy review shows, ChatGPT isn’t just an AI chatbot. It’s a bundle of AI tools that each target completely different tasks.

That’s actually one of ChatGPT Pro’s greatest weaknesses compared to alternatives. You can cobble together multiple competing AI services that are each optimized for their particular tasks, and it’ll likely cost significantly less. This piecemeal approach does mean managing more subscriptions, but it feels more practical than OpenAI’s strategy of bundling everything into one flat subscription fee.

To be fair, ChatGPT Pro does offer some features that you can’t get elsewhere or don’t live up to OpenAI’s quality. Deep Research is especially compelling if you want an AI research assistant at your beck and call. But the other stuff, like OpenAI’s Operator agent, aren’t reliable enough yet to deliver enough value for the price.

As of this writing, the best reason to get ChatGPT Pro is to boost usage rate limits. The rate limits with ChatGPT Pro are high enough that they’re virtually non-existent. Competing services, like Anthropic’s Claude or DeepSeek, can get bogged down at peak times and will throttle you even if you’re a paid subscriber.

Even so, the exorbitant price of ChatGPT Pro is hard to justify unless you’re hitting its AI services with hundreds of requests every day. And if that’s the case, you might be better off accessing OpenAI’s models directly through its API services (which charge you per millions of tokens instead of a flat monthly fee).

For most people, ChatGPT Plus should suffice. Otherwise, if you’re a power user, professional, or enthusiast looking to benefit from the latest AI advancements, I recommend subscribing to multiple different AI services that each target more specific use cases rather than adopting ChatGPT Pro’s all-in-one package approach.