Gemini Free Review: The Best Free AI Chatbot I’ve Used

Google Gemini has come a long way. after that Previous life as a bardwhere the inaccuracies of the hallucinations have caused Google to fall, Brand as Geminiwhere is it? Hallucination inaccuracies resulted in Google’s stock price falling. Two years after in the oven, Google seems to have finally got things right.

The free version of Gemini, running on the 2.5 flash model, is a powerful product. It’s quick to answer questions and the facts are correct. In the case of Gemini, we found that there is a higher rate limit when compared to the free version of ChatGPT. One test allows you to generate multiple images in Gemini, chatgpt freeone image throws me across my limits. In our test, ChatGpt’s image generator was the slowest CNET tested. Image generation was much faster than ChatGpt Free. At the same time, I even slammed Gemini’s token limit just by asking the first question of the day.

For casual AI users, Gemini 2.5 Flash is enough. This will not give users sufficient access to competent AI models and make them feel at the edge of constantly reaching limits. Still, the model is not perfect and can make mistakes, especially on the imaging side of things. Nevertheless, for a typical use case, the free version of Gemini can take over Google search.

How CNET reviews AI models

This year we took a different approach to reviewing AI chatbots. Simpler queries do not emphasize the model as the AI ​​model has been improved. These models are also connected to the Internet, which helps to ensure accuracy. Instead, I took a more experiential approach. AI chatbots are all machines. The way you use AI chatbots as a writer and journalist is different from coders, lawyers and artists. Thankfully, since journalists are generalists, I feel that my use applies to a wide range of users. This means asking the exact same question for all the AI ​​models you use and simply not comparing the answers.

How accurate is Gemini?

Compared to past iterations of Gemini, Gemini Free, running on a newly updated 2.5 Flash model, is almost very accurate. This is because Gemini not only has an open connection to the Internet to cross-reference information, but 2.5 Flash is the “thinking” model. What this means is that the model does not simply act as an “autocomplete of steroids.” Before giving output, you try to follow a set of rules and reasons.

Of course, there is a lot of debate as to whether thought and inference models are actually doing it, and whether it is actually only an increase in mathematical calculations that gives greater accuracy to sentence generation. Anyway, with Gemini 2.5 flash you can actually see how the model thinks. Deepseek R1It was a hit on the internet at the end of last year.

As the new owner of the Nintendo Switch, I’m looking forward to seeing which games will come to the console. Rumors have been raised that the PS5-only (with a recent PC release) will be two star blades that will be coming to Switch 2. Given that it’s a technically demanding game, I was curious to see how it would run on Nintendo’s new handheld.

The Gemini 2.5 Flash did an amazing job. I’ve seen how well the Unreal Engine 4 titles run on the original switch and how well they run on the more powerful switch to each other. We found that in docking mode it’s most likely to run at 1080p at a consistent 30 fps using AI upscaling called DLSS. I’m not a hardware expert, but this conclusion seems to match with other switch 2 ports like the Cyberpunk 2077.

We have also looked into whether it is wise or economic to do Turo in New York City. Turo is a car rental service that allows individuals to rent cars as a way to make money. Think of it like Airbnb. New York can be a challenging market given parking restrictions, street rules and other costs of ownership. The Gemini 2.5 Flash did an amazing job of breaking the reason why renting a manual transmission Toyota GR86 can benefit the TURO as a more enthusiastic vehicle, or is it a niche for most drivers.

Gemini also pointed out certain engine issues with the 2022 model. This is something to consider, although rare. After that, I broke the math and the kind of revenue and profit I might be seeing. It gave me a low, medium and high estimate. Renting a manual transmission Toyota GR86 helped us to conclude that it could lead to more trouble than it was worth it. I think Gemini was able to pull from both the dedicated forum and Reddit data. (Google signed a $60 million licensing agreement with Reddit last year. )

Gemini can be pulled from YouTube, Google Maps and a variety of other Google-owned products. This gives Gemini an advantage over other AI chatbots. For example, if you want to know what ingredients a restaurant uses in burritos, Gemini can cross-reference Google Maps reviews to find the answer. For example, ChatGpt needs to search Yelp and other resources to find the answer.

Rate limiting is rare and random

Google says Newly updated Gemini 2.5 flash model There is a million token context windows. This is far beyond even the paid version of ChatGPT’s GPT-4O model, which has a 128,000 tokens. Certainly, CHATGPT’s flagship GPT-4.1 model has a million token context windows.

When using Gemini 2.5 Flash I didn’t encounter many limitations. I continued asking the question and it allowed me to generate multiple images without restricting me.

But randomly, I run into my limits, even if it was my first question of the day. When that happened I had to wait a few hours for it to reset. I don’t know how Google measures its usage. Is it based on how much you use it in an hour or does it accumulate over several days? The latter certainly doesn’t make sense.

Given that Google says that Gemini has a million token context windows, I was surprised when I couldn’t paste it into the transcript from the 2 hour meeting for a summary. Oddly, I could only paste one quarter of that. When I asked Gemini why I couldn’t paste any more, I assert that I could do that big context window that I was sure could handle anything I could throw. I tried it again. Same result.

After uploading the .txt file I was able to read and summarize the entire meeting. When asked Gemini why, he said that Google could put character restrictions on direct text input to prevent the browser from slowing down. I did not run into this issue with the paid version of Gemini.

I didn’t test that coding feature, but Google says Gemini Code Assist It offers 180,000 completions per month to free users. This means that users must code 14 hours a day per day before they can meet the limits, according to the company.

Google is strangely late to shopping compared to ChatGpt

Google makes money from online ads. 78% of revenue in 2024. Google searches aren’t just filled with recent ads, but in my opinion, product carousels and sponsored products posting can be offensive. Still, we are hoping that Google’s AI chatbot will become a shopping powerhouse, right?

For shopping, the gemini 2.5 flash is much behind Chatgpt. Early this year, Openai has issued an update for all ChatGPT users Links to products and corresponding images to make shopping a dynamic experience within the chatbot. I found it ChatGpt Free Shopping Experience It’s pretty good despite occasional linking hiccups. Openai says it doesn’t monetize its shopping recommendations.

But shopping in Gemini is an inactive experience. Certainly, for product research, Gemini 2.5 Flash can lift the information you need and cross-compare products. However, unless you ask, it will not link to the product. Also, it doesn’t pull out images like Google searches.

When purchasing a webcam to connect to Mario Kart World Gaming’s Nintendo Switch 2, Gemini did a solid job of recommending the product and even cross-referenced the linked Reddit Post.

Oddly, when I asked Gemini 2.5 Flash about the webcam recommendations for Nintendo Switch 2, I ran into an error saying, “Something went wrong.” There was no explanation. I waited a bit, but eventually I had to start a new chat for things to start working again.

Image Generation with Gemini Free: You Get What You Pay

Gemini 2.5 flash is extremely generous in image generation. Unfortunately, generating the correct image is an annoying process.

I wanted Gemini to create a nostalgic image of him playing a boy playing a game boy behind his parents’ car during a night road trip. Gemini was able to create images, but the logic of the world was completely off.

Initially, Gemini 2.5 Flash produced images of sad boys.

When I called Gemini, when I said this wasn’t what I was looking for, it was a course revised but still didn’t do a great job. One of the subsequent attempts was certainly excellent, but there was no color palette I was looking for. The boy was also in the front seat of the car. It’s really not safe. In another iteration, the background car drove off, which does not match the correct world logic.

After the exchange, Gemini continuously produced the wrong images, which looked strange. For example, the boy is currently in the front seat with his parents, but is heading backwards. I finally gave up.

The Gemini 2.5 Flash is fast and generous in its image generation, but it’s far from ideal. Google needs to work on fixing internal logic within Gemini. Demis Hassabis of Google Deepmind spoke about the “world model” on Google I/O earlier this year. These models did a better job of understanding and expressing realistic physics. Hopefully, this technology will soon drip down to the free version of Gemini.

Gemini Free offers more than ChatGpt for free

Google deserves credit for how much improvements have been made with Gemini over the past year. AI chatbots offer a much more accurate and feature-rich experience. It is also impressive that many of this is given free. It will definitely notify other AI chatbots. It shows how Google’s power is difficult to compete, especially when a company is trying to establish Gemini as a people’s AI chatbot.

ChatGpt, Claude and Perplexity can compete by providing higher quality and more accurate information. Of course, this will require more investment, innovation and server costs. This may be more difficult for businesses that are less abundant than Google. For example, the free version of Claude runs on Sonnet 4, a “hybrid inference model,” and uses a multi-layer approach to make the most of AI.

Still, it’s impressive that Google offers everyone access to a zero dollar “thinking” model. Considering that Deepseek R1 did the same thing earlier this year, this could have forced Google. Anyway, this new and improved Gemini is a step in the right direction.