The best large language models (LLMs) in 2025

Large language models (LLMs) are the main kind of text-handling AIs, and they're now absolutely everywhere. 

ChatGPT is the most famous tool that openly uses an LLM, but Google uses one to generate AI answers in Search, and Apple launched the LLM-powered Apple Intelligence on its devices late last year. And that's before you consider any of the other chatbots, text generators, and other tools built on top of LLMs.

LLMs have been studied in research labs since the late 2010s, but after the release of ChatGPT (which showcased the power of GPT), they've burst out of the lab and into the real world.

We're now into the fifth and sixth generations of LLMs, and with that, they're increasingly useful and powerful. Reasoning models that take extra time to work through hard problems are the biggest shift in the past year. Before that, it was the introduction of large multimodal models (LMMs), which are able to handle other input and output modalities, like images, audio, and video, as well as text. Of course, the rapid pace of these developments complicates things even more. So here, I'll break down some of the most important LLMs, LMMs, and reasoning models on the scene right now.

• The best LLMs

• What is an LLM?

• What is an open source LLM?

• How do LLMs work?

• What are reasoning models?

• What can LLMs be used for?

• Why are there so many LLMs?

• What to expect from LLMs in the future

The best LLMs

There are dozens of major LLMs, and hundreds that are arguably significant for some reason or other. Listing them all would be nearly impossible, and in any case, it would be out of date within days because of how quickly LLMs are being developed.

(I'm updating this list now for the first time in a few months, and there are already plenty of new versions of models to talk about—plus a few new models to add. The last time I updated the list, o1 hadn't even been announced; now we have o3!)

Take the word "best" with a grain of salt here: I've tried to narrow things down by offering a list of the most significant, interesting, and popular models, not necessarily the ones that outperform on benchmarks (though most of these do). I've also mostly focused on LLMs, LMMs, and reasoning models that you can actually use—rather than ones that are the subjects of super interesting research papers or just teased in marketing materials—since we like to keep things practical around here.

Click on any app in the list below to learn more about it.

What is an LLM?

An LLM, or large language model, is a general-purpose AI text generator. It's what's behind the scenes of all AI chatbots, AI writing generators, and most other AI-powered features like summarized search answers.

Stripped of fancy interfaces and other workarounds, LLMs take a prompt and generate an answer. The chatbots built on top of LLMs aren't looking for keywords so they can answer with a canned response—instead, they're doing their best to understand what's being asked and reply appropriately.

This is why LLMs have really taken off: the same models (with or without a bit of extra training) can be used to respond to customer queries, write marketing materials, summarize meeting notes, and do a whole lot more.

But LLMs can only work with text, which is why LMMs are increasingly popular: they can incorporate images, handwritten notes, audio, video, and more. Many of the biggest models are now LMMs.

What is an open source LLM?

There are three major categories of LLM: proprietary, open, and open source.

Proprietary models like GPT-4o and Claude 3.5 are some of the most popular and powerful models available, but they're developed and operated by private companies. The source code, training strategies, model weights, and even details like the number of parameters they have are all kept secret. The only ways to access these models are through a chatbot or app built with them, or through an API. You can't just run GPT-4o on your own server.

Open and open source models are more freely available. You can download Llama 3, Gemma 2, and DeepSeek R1 from Hugging Face and other model platforms and run them on your own devices—and even re-train them with your own data to create your own model. Developers can build their own chatbots and apps on top of them. You can even dig deep into things like the model weights and system architecture to understand how they work (as best as anyone can).

So what's the difference between open and open source? Well, companies like Meta and Google say that Llama 3 and Gemma 2 are open as though it's the same as open source, but there's a major distinction. 

Open source licenses are incredibly permissive. Mostly, you have to agree to make anything you build with it open source as well—and give attribution to the original developers. If you want to build a multi-billion dollar company off open source software or create a crime chatbot that tells people how to rob a bank, you're absolutely free to do so. The police might have some issues with the latter project, but you wouldn't be breaking any software licenses. 

Open licenses are still permissive, but they have some additional limits. For example, Llama 3's license allows commercial use up to 700 million monthly users and blocks certain uses. You or I could build something with it, but Apple and Google can't. Similarly, Gemma 2's prohibited use policy, among other things, bans "facilitating or encouraging users to commit any type of crimes." Understandably, Google doesn't want to see unsavory bots "powered by Google Gemma" plastered all over the news.

How do LLMs work?

Early LLMs, like GPT-1, would fall apart and start to generate nonsense after a few sentences, but today's LLMs, like GPT-4o, can generate thousands of words that all make sense.

To get to this point, LLMs were trained on huge corpuses of data. The specifics vary a little bit between the different LLMs—depending on how careful the developers are to fully acquire the rights to the materials they're using—but as a general rule, you can assume that they've been trained on something like the entire public internet, every book, newspaper, and magazine that's ever been published, and the synthetic output of earlier AI models at a bare minimum. This is why LLMs can generate text that sounds so authoritative on such a wide variety of subjects.

From this training data, LLMs are able to model the relationship between different words (or really, fractions of words called tokens) using high-dimensional vectors. This is all where things get very complicated and mathy, but the basics are that every individual token ends up with a unique ID and that similar concepts are grouped together. This is then used to generate a neural network, a kind of multi-layered algorithm based on how the human brain works—and that's at the core of every LLM. 

The neural network has an input layer, an output layer, and multiple hidden layers, each with multiple nodes. It's these nodes that compute what words should follow on from the input, and different nodes have different weights. For example, if the input string contains the word "Apple," the neural network will have to decide to follow up with something like "Mac" or "iPad," something like "pie" or "crumble," something like "by Charli XCX," or something else entirely. When we talk about how many parameters an LLM has, we're basically comparing how many layers and nodes there are in the underlying neural network. In general, the more nodes, the more complex the text a model is able to understand and generate.

LMMs are even more complex because they also have to incorporate data from additional modalities, but they're typically trained and structured in much the same way.

Of course, an AI model trained on the open internet with little to no direction sounds like the stuff of nightmares. And it probably wouldn't be very useful either, so at this point, LLMs undergo further training and fine-tuning to guide them toward generating safe and useful responses. One of the major ways this works is by adjusting the weights of the inputs and outputs of different nodes, though there are other aspects of it too.

All this is to say that while LLMs are black boxes, what's going on inside them isn't magic. Once you understand a little about how they work, it's easy to see why they're so good at answering certain kinds of questions. It's also easy to understand why they tend to make up (or hallucinate) random things.

What are reasoning models?

Reasoning models like OpenAI o3 and DeepSeek R1 are LLMs that are trained to generate a response using Chain-of-Thought (CoT) reasoning. 

When they're given a prompt, instead of replying as quickly as possible, they break the problem down into multiple simple steps and attempt to work through them. If they encounter issues, they can reassess and approach problems from a different angle.  

This kind of reasoning requires more computing resources, but it tends to lead to more powerful AI models. 

What can LLMs be used for?

LLMs are powerful mostly because they're able to be generalized to so many different situations and uses. The same core LLM (sometimes with a bit of fine-tuning) can be used to do dozens of different tasks. While everything they do is based around generating text, the specific ways they're prompted to do it changes what features they appear to have. 

Here are some of the tasks LLMs are commonly used for:

• General-purpose chatbots (like ChatGPT and Google Gemini)

• Summarizing search results and other information from around the web

• Customer service chatbots that are trained on your business's docs and data

• Translating text from one language to another

• Converting text into computer code, or one language into another

• Generating social media posts, blog posts, and other marketing copy

• Sentiment analysis

• Moderating content

• Correcting and editing writing

• Data analysis

And hundreds of other things. We're only in the early days of the current AI revolution.

But there are also plenty of things that LLMs can't do, but that other kinds of AI models can. A few examples:

• Interpret images

• Generate images

• Convert files between different formats

• Create charts and graphs

• Perform math and other logical operations

Of course, some LLMs and chatbots appear to do some of these things. But in most cases, there's another AI service stepping in to assist—or you're actually using an LMM.

With all that context, let's move on to the LLMs themselves.

The best LLMs

GPT-4o

• Developer: OpenAI

• Parameters: More than 175 billion

• Context window: 128,000

• Access: API

OpenAI's Generative Pre-trained Transformer (GPT) models kickstarted the latest AI hype cycle. There are two main models currently available: GPT-4o and GPT-4o mini. Both are also multimodal models, so they can also handle images and audio.

All the different versions of GPT are general-purpose AI models with an API, and they're used by a diverse range of companies—including Microsoft, Duolingo, Stripe, Descript, Dropbox, and Zapier—to power countless different tools. Still, ChatGPT is probably the most popular demo of its powers.

OpenAI plans to release a more powerful version called GPT-4.5 later this year.

You can automate the power of GPT by connecting it to Zapier, so you can use GPT straight from the other apps in your tech stack. Here's more about how to automate ChatGPT, or you can get started with one of these pre-made workflows.

o3 and o1

• Developer: OpenAI

• Parameters: More than 175 billion

• Context window: 200,000

• Access: API

OpenAI's o1 model was the first reasoning model. Since it launched, OpenAI's reasoning models have topped almost every benchmark and head-to-head test. So far, there's been o3-mini, o1, o1-preview, and o1-mini, with o3 due later this year.

Like GPT-4o, o1 and o3-mini are available through an API and ChatGPT. Fortunately, OpenAI has recognized that all these different AI models are getting confusing and that the naming systems are terrible and incompatible. At some point in the future, it will release GPT-5, which will combine the GPT and reasoning models into one line.

You can also connect Zapier to OpenAI's reasoning models through ChatGPT, so you can use o1 or o3-mini straight from the other apps in your tech stack.

Gemini 

• Developer: Google

• Parameters: Nano available in 1.8 billion and 3.25 billion versions; others unknown

• Context window: Up to 2 million

• Access: API

Google Gemini is a family of AI models from Google. The six models—Gemini 1.0 Nano, Gemini 2.0 Flash, Gemini 2.0 Flash Thinking, Gemini 2.0 Flash-Lite, Gemini 2.0 Pro, and Gemini 1.0 Ultra—are designed to operate on different devices, from smartphones to dedicated servers, and cover a wide variety of uses. Gemini 2.0 Flash Thinking is Google's first reasoning model.

While capable of generating text like an LLM, the Gemini models are also natively able to handle images, audio, video, code, and other kinds of information. They're optimized for a long context window, which means they can process larger volumes of text.

Gemini 2.0 Pro and 2.0 Flash also power AI features throughout Google's apps, like Docs and Gmail, as well as Google's chatbot, which is confusingly also called Gemini. Google's Gemini models are available to developers through Google AI Studio or Vertex AI.

With Zapier's Google Vertex AI and Google AI Studio integrations, you can access Gemini from all the apps you use at work. Here are a few examples to get you started, or you can learn more about how to automate Google AI Studio.

Gemma

• Developer: Google

• Parameters: 2 billion, 9 billion, and 27 billion

• Context window: 8,200

• Access: Open

Google Gemma is a family of open AI models from Google based on the same research and technology it used to develop Gemini. The latest version, Gemma 2, is available in three sizes: 2 billion, 9 billion, and 27 billion parameters.

Llama

• Developer: Meta

• Parameters: 1 billion, 3 billion, 8 billion, 11 billion, 70 billion, 90 billion, and 405 billion

• Context window: 128,000

• Access: Open

Llama 3 is a family of open LLMs from Meta, the parent company of Facebook and Instagram. There are currently seven models split between Llama 3.1, Llama 3.2, and Llama 3.3. 

In addition to powering most AI features throughout Meta's apps, the Llama "herd" is one of the most popular and powerful open LLM families, and you can download the source code yourself from GitHub. Because it's free for research and commercial uses, a lot of other LLMs use a Llama model as a base.

Currently, the models in the lineup are the text-only Llama 3.1 8B, Llama 3.1 405B, and Llama 3.3 70B, and the multimodal Llama 3.2 1B, Llama 3.2 3B, Llama 3.2 11B, and Llama 3.2 90B.

R1

• Developer: DeepSeek 

• Parameters: 671 billion

• Context window: 128,000

• Access: Open, chatbot, API

DeepSeek R1 caused a major stir when it launched. It's a reasoning model that's as capable as OpenAI o1, but it was developed by a Chinese tech company using more limited computer hardware on a far smaller budget and released as an open model. 

Despite this impressive achievement, the full implications of DeepSeek's compute-saving innovations haven't yet been realized. It also remains to be seen how future sanctions will affect the AI company. 

V3

• Developer: DeepSeek 

• Parameters: 671 billion

• Context window: 128,000

• Access: Open, chatbot, API

DeepSeek V3 is DeepSeek's equivalent of GPT-4. It's a state-of-the-art open LLM without reasoning or multimodal features. Like R1, it was developed using more limited computer hardware and for less financial investment than typical LLMs.

Similarly to R1, while V3 is an impressive technological achievement, it remains unclear how popular it will prove over the next year or two. 

Claude

• Developer: Anthropic

• Parameters: Unknown 

• Context window: 200,000

• Access: API

Claude is arguably one of the most important competitors to GPT. Its three models—Claude 3.7 Sonnet, Claude 3.5 Haiku, and Claude 3 Opus—are designed to be helpful, honest, harmless, and crucially, safe for enterprise customers to use. As a result, companies like Slack, Notion, and Zoom have all partnered with Anthropic.

Like all the other proprietary LLMs, Claude is only available as an API, though it can be further trained on your data and fine-tuned to respond how you need. You can also connect Claude to Zapier so you can automate Claude from all your other apps. Here are some pre-made workflows to get you started.

Command

• Developer: Cohere 

• Parameters: Command R7B has 7 billion; the other models are unknown

• Context window: Up to 128,000

• Access: API

Like Claude 3, Cohere's Command models are designed for enterprise users. Command R7B, Command R and Command R+ offer an API and are optimized for retrieval augmented generation (RAG) so that organizations can have the model respond accurately to specific queries from employees and customers.

As a result, companies like Oracle, Accenture, Notion, and Salesforce use Cohere's models.

Nova

• Developer: Amazon 

• Parameters: Unknown

• Context window: Up to 300,000

• Access: API

Amazon Nova is a family of frontier models available on Amazon Web Services. Despite a slow start from Amazon, the current models—including Amazon Nova Micro, Amazon Nova Lite, and Amazon Nova Pro—perform competitively in a range of benchmarks. Given AWS's prominence in cloud computing, Amazon Nova models may end up proving popular.

Mistral Large 2

• Developer: Mistral

• Parameters: 123 billion

• Context window: 128,000

• Access: Open weight 

Mistral is one of the largest European AI companies. Its Mistral Large 2 model, Pixtral Large multimodal model, and Le Chat chatbot are all direct competitors to GPT-4o, Gemini, ChatGPT, and other state-of-the-art AI tools. 

Mistral Large 2 is available with open weights for research and commercial purposes. While this isn't a fully open license, it means it can be trained for specific purposes. 

Qwen

• Developer: Alibaba Cloud 

• Parameters: 0.5 billion, 1.5 billion, 3 billion, 7 billion, 14 billion, 32 billion, and 72 billion

• Context window: Up to 1,000,000

• Access: Open, API, chatbot

Qwen is a family of AI models from Chinese tech giant Alibaba. There are dozens of open models available across the different Qwen2.5 families, including models tailored for vision, coding, math, and a million-token context.

The highest performing model, Qwen2.5 Max, matches or exceeds models like GPT-4o and Gemini 2.0 Pro across a wide range of benchmarks. Right now, it's only available through the Qwen chatbot and API.

Phi-3 and Phi-4

• Developer: Microsoft

• Parameters: 3.8 billion, 7 billion, and 14 billion

• Context window: Up to 128,000

• Access: Open 

Microsoft's Phi-3 family of small language models are optimized for performance at small size. The 3.8 billion parameter Mini, 7 billion parameter Small, 14 billion parameter Medium, and 14.7 billion parameter Phi-4 all out perform larger models on language tasks.

The models are available through Azure AI Studio, Hugging Face, and other open model platforms.

Grok

• Developer: xAI

• Parameters: Unknown

• Context window: 128,000

• Access: Chatbot and open

Grok, an AI model and chatbot trained on data from X (formerly Twitter), originally didn't warrant a place on this list on its own merits. Grok 3, however, offers state-of-the-art performance and reasoning abilities. In one benchmark, it outperforms every other model.   

Still, while its performance now matches other models, I'm mostly listing it here because it was developed by xAI, the AI company founded by Elon Musk. Its stay at the top of any benchmark isn't likely to be long, and it isn't making big waves in the AI scene, but xAI is going to continue to get plenty of media coverage, so it's worth knowing it exists.

Read more in our comparison of Grok vs. ChatGPT.

Why are there so many LLMs?

Until a few years ago, LLMs were limited to research labs and tech demos at AI conferences. Now, they're powering countless apps and chatbots, and there are hundreds of different models available that you can run yourself (if you have the computer skills). How did we get here?

Well, there are a few factors in play. Some of the big ones are:

• With GPT-3 and ChatGPT, OpenAI demonstrated that AI research had reached the point where it could be used to build practical tools—so lots of other companies started doing the same. 

• LLMs take a lot of computing power to train, but it can be done in a matter of weeks or months.

• There are lots of open models that can be re-trained or adapted into new models without the need to develop a whole new model.

• There's a lot of money being thrown at AI companies, so there are big incentives for anyone with the skills and knowledge to develop any kind of LLM to do so.

What to expect from LLMs in the future

I think we're going to continue to see a lot of innovation and new LLMs in the near future. DeepSeek was able to launch two state-of-the-art models using significantly less powerful hardware. On the back of that, I feel the barrier to developing a powerful AI model has never been lower. Major tech companies like Apple, Amazon, IBM, Intel, and NVIDIA also have compelling reasons to continue developing their own LLMs, even if they're only used internally. 

I also think we're going to continue to see more efficient LLMs tailored to run on smartphones and other lightweight devices. Google started this with Gemini Nano, and while Apple Intelligence has attracted some criticism, it still runs natively on edge computing devices. 

And, of course, every major AI provider will launch a model with reasoning. So far, the major options are OpenAI and DeepSeek. But Google, Claude, Cohere, and Meta will presumably debut full reasoning models in 2025.

Other than that, who can tell? Three years ago, I definitely didn't think we'd have powerful AIs like ChatGPT available for free. Maybe in a few years, we'll have artificial general intelligence (AGI).

Related reading:

• The best AI productivity tools

• The best AI courses for beginners

• Claude vs. ChatGPT: What's the difference?

• How to use Claude Artifacts for data visualization

This article was originally published in January 2024. The most recent update was in February 2025.