Ultimate guide to ChatGPT, Gemini, Llama, and other genAI chatbots you need right now

Since OpenAI first released ChatGPT to the public in November 2022, there has been a proliferation of consumer-facing chatbots in the large language model (LLM) race, changing how consumers interact with information. 

Chatbots are the interface we use to interact with LLMs. These models are trained on vast data sets and fine-tuned for conversational abilities and are constantly improving. LLMs now power a new generation of chatbots that assist with everything from writing essays and summarizing documents to generating complex code. 

While many LLMs are designed for niche applications or B2B purposes, some chatbots cater specifically to the general user, offering tools you can use today, for free or through a subscription. 

From startups that have quickly become household names like OpenAI and Anthropic, to legacy tech companies like Google and Meta, here is a guide to some of the most useful LLM-powered chatbots for everyday use, from drafting emails to analyzing data.

ChatGPT

ChatGPT is a chatbot developed by OpenAI, a leading AI research lab initially founded in 2015 as a nonprofit. Since CEO Sam Altman’s brief ouster in November 2023, nearly the entire original leadership team has departed, some warning they no longer believe OpenAI will build superintelligence responsibly. Now, OpenAI is attempting a transition to a reportedly for-profit public benefit corporation, like rival Anthropic.

ChatGPT first launched in November 2022 as a free research preview using the GPT-3.5 model and quickly amassed over one million users within five days. As of October 2024, it has 200 million weekly active users worldwide. 

Users can access ChatGPT through the website or app. To query ChatGPT, type a prompt, such as “What are the health benefits of strawberries?” in the message bar of the homepage and click the Send message icon (or hit Enter). The chatbot will provide a response, and users can continue the conversation or edit queries further. 

The chatbot can summarize a PDF, debug code, and can browse the web, enabling real-time internet searches to provide up-to-date, accurate information. It can also recall earlier conversations, enabling it to better understand the context of requests. Like all models, the more specific the prompt is, and the more context is given, the stronger the output will be. 

On the app, ChatGPT’s Advanced Voice Mode, which has been free since October, allows users to have a verbal conversation with the chatbot. Users initiate the conversation by pressing the voice icon and speaking in any language, and can use the tool in a myriad of ways, including simulating conversations with historical figures, receiving a personalized virtual museum tour, or practicing a new language.

In May 2024, OpenAI introduced GPT-4o, its most advanced model to date. GPT-4o accepts any combination of text, audio, image, and video as input, and in return can generate any combination of text, audio, and image as an output. This multimodal capability means users can upload photos and PDFs for ChatGPT to analyze, and its average response time to audio inputs is 320 milliseconds, similar to human response time⁠ in a conversation.

OpenAI also launched GPTs this past November, allowing users to customize ChatGPT for specific purposes without any coding. Users can make their own through giving instructions and examples, and can also integrate external databases or interact with the real world by making APIs available to the GPT. 

The GPT Store features creations by verified builders, and OpenAI spotlights useful GPTs in categories like productivity, research and analysis, and programming.  

The company also launched a feature called Canvas in October 2024. When working on writing or coding projects that go beyond a simple chat, Canvas is a new interface that automatically launches in a separate window, allowing users to collaborate with the system on a project. 

“Like a copy editor or code reviewer, it can give inline feedback and suggestions with the entire project in mind,” OpenAI wrote in the feature’s press release. ChatGPT 4o with Canvas is available in beta to some paying subscribers, and OpenAI plans to roll it out to all users. 

In September 2024, the company released its newest model GPT o1, specifically trained to excel in complex tasks in science, coding, and math. While it excels at questions requiring logical rigor, like debugging extensive codebases, or solving intricate math problems, it comes at the cost of longer compute times, compared to models like GPT-4o. It also lacks many features that GPT-4o has, including memory functions, file upload capabilities, data analysis tools, and web browsing abilities. 

Claude 

Anthropic was founded in 2021 by several ex-OpenAI employees, and positions itself as an AI safety first company. Its latest LLM is the Claude 3.5 Sonnet, which currently powers Anthropic’s chatbot named Claude.  

The Claude 3.5 model has established itself as a leader in agentic coding, with leading scores in coding benchmarks, and a long context capability. It is multimodal, but unlike competitors, does not have real-time access to the internet. 

Anthropic’s Constitutional AI approach means it embedded predefined ethical guidelines directly into the AI’s operational framework, impacting the entirety of its decision-making processes. These guidelines are inspired by human rights documents, ethical codes, and best practices in AI development, spanning fairness, nondiscrimination, transparency, and accountability.

Claude’s Artifacts feature allows the chatbot to produce stand-alone content like code snippets, documents (Markdown or Plain Text), data visualizations, and even websites (single-page HTML). Artifacts appear in a separate window or panel within the conversation, making it easier to view, copy, and interact with more complex content, similar to OpenAI’s Canvas feature. 

Claude automatically creates an Artifact for significant, self-contained content over 15 lines that is complex, reusable, and useful for editing or reference outside the conversation.

Claude.ai Pro and Team subscribers have also had access to another feature called Projects since June 2024. Projects enables users to create specialized workspaces that organize documents, code snippets, and other contextual information that Claude can reference during conversations within that Project.

Gemini 

Gemini is a chatbot created by Google that essentially replaced its previous chatbot named Bard. It’s based on an LLM also called Gemini, which was developed by DeepMind, now a part of Google. 

In May 2024, Google introduced Gemini 1.5 Pro, the first in what the company called a “mid-sized multimodal model.” It also released Gemini 1.5 Flash, a faster version of Gemini mainly for developers, according to The Verge. 

Gemini has real-time access to the internet and can use Google Drive to summarize docs and PDFs, get quick answers, and find information in content. Its responses can be exported to a Gmail, Doc, or Sheet. For example:

• Brainstorming: Get ideas for professional development or lesson plans
• Write: Create first drafts of emails and blog posts 
• Summarize: Summarize PDFs, meeting notes, and long email threads
• Create images: Generate images on the fly 
• Plan: Make plans with Google Maps and Google Flights 
• Code: Find bugs, syntax errors, and logical errors in code 
• Get directions: Get suggestions on nearby stores, restaurants, businesses, and landmarks 

Google is integrating Gemini into its existing suite of apps and creating new features, like the AI-generated responses at the top of Google searches, and a new “Ask Photos” feature on Google Photos. 

While Gemini is closed source, Google is also building Gemma, a collection of lightweight, more open-source models designed mainly for developers and researchers. 

Llama 

Meta announced the Meta AI chatbot in September 2023. It is currently powered by LlaMA 3.0, and can also process and generate multimodal content, as well as access real-time information across the web. 

Llama 3 has officially been integrated into Meta applications like Instagram, Facebook, WhatsApp, and Messenger. Users can use the chatbot inside the applications to do a variety of tasks:

• Create content: Generate different types of written content, from fictional stories to data-driven reports
• Draft emails: Craft the right response and maintain a consistent tone across channels. 
• Analyze data: Llama 3 can summarize data findings or documents to generate visual reports for better decision-making. 
• Generate code: Developers can generate code snippets, identify bugs, or see programming recommendations to improve processes. 

Due to the way that Llama 3 is integrated into the applications, it currently cannot be disabled by the user, and usage will help inform training the model. 

Mistral 

French AI startup Mistral has built the multimodal chatbot Le Chat, which can search the web, and has a Canvas tool similar to ChatGPT.

It’s multimodal, and the image generator it currently uses, Black Forest Labs Flux Pro 1.1, outperforms all other image generators in terms of quality and inference speeds, according to Decrypt.

Mistral’s underlying models are fueling the chatbot’s capabilities. Pixtral Large was also released in November, and rivals Anthropic’s Claude 3.5 Sonnet, Google’s Gemini 1.5 Pro, and OpenAI’s GPT-4o on certain multimodal benchmarks. 

“Particularly, Pixtral Large can understand documents, charts, and natural images,” according to Mistral’s blog post. “The model demonstrates frontier-level image understanding.”

Mistral also unveiled a new version of Mistral Large, a flagship line of text-only models, one day after Meta dropped Llama 3.1 this past November. According to TechCrunch, it outpaces Llama 3.1 405B on code generation and math performance, with less than a third of the parameters. 

Mistral said the team focused on enhancing the models’ reasoning capabilities, minimizing its tendency to hallucinate, and training it to acknowledge when it cannot find solutions or has insufficient information to provide a confident answer. 

Mistral introduced an early version of Le Chat Agents in November, where users can codify workflows as agents to automate them. With instructions and examples on Le Chat or through the API, users can wrap models with additional context and instruction to create custom workflows for repetitive tasks such as receipt scanning for expense reporting, summarizing meeting notes, or processing invoices.

It can also provide insights for business intelligence including:

• Analyzing data to extract insights: Can process large data sets efficiently, helping reveal trends and patterns.
• Predictive maintenance: Detect equipment failures and maintenance needs to minimize downtime and reduce maintenance costs.

As the proliferation of LLMs continues, companies are pushing boundaries to integrate these tools into everyday life, making them more accessible, versatile, and capable than ever.

OpenAI’s ChatGPT has set the standard for versatility and customization, while Anthropic’s Claude emphasizes safety and ethical considerations. Google’s Gemini and Meta’s Llama integrate their AI chatbot features into their ecosystems. Meanwhile, startups like Mistral are innovating with specialized tools like Agents.

As companies continue refining their LLMs, consumer-facing chatbots are poised to play an increasingly central role in how we interact with information and technology. These chatbots have some form of free access and are all available right now.

https://www.fastcompany.com/91240580/ultimate-guide-to-chatgpt-gemini-llama-and-other-genai-chatbots-you-need-right-now?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss