The productive AI landscape has turned into a high-stakes battleground in 2024, with an army of upstarts storming the citadel that OpenAI once ruled.

It seems like everyone and their tech-savvy grandma are vying for a piece of the AI ​​pie, preparing language models, agent AI, image generators, and even a meme AI coin or two.

Standards are changing faster than our human ability to keep up. Hardly a week goes by without a shiny new toy hitting the market – an updated LLM course here, a turbocharged image generator there, or a next-generation AI showcasing some weird training technique.

But here in Decryptionwe've rolled up our sleeves and tried them all.

We've kicked the tires, pushed the buttons, and delved into the inner workings and outputs provided by the most popular AI models, some of which are less well known.

Now that it's clear that OpenAI isn't the only mayor in town, we've compiled a list of the cream of the crop — the generative AI models that have dazzled us, confounded us, and occasionally made us spit out our coffee.

Chat bots

A chatbot is a computer program designed to simulate conversation with human users. It uses natural language processing and artificial intelligence to understand user inputs and create appropriate responses. Usually, people confuse chatbots with LLMs, or large language models.

Today, chatbots have become a bit more sophisticated, with capabilities that go beyond just generating text. They can now browse the web, create and understand images, talk to the user, etc.

Here is our list of the best chatbots you should try:

Gold Medal: ChatGPT from OpenAI

ChatGPT offers a wide range of features at $20 per month, including custom agent creation using natural language, a clean interface, web search, and multiple models (inference, typing, vision, audio, and image generation).

Silver Medal: Claude Anthropy

Premium LLM With an intuitive user interface featuring split-screen elements for inference and code generation, Claude supports 1 million token context and custom agents. However, it lacks web search and image generation and often has capacity issues, forcing users to switch to a weaker model or create shorter, “summary” answers. For this reason, it can't be the best yet.

Bronze: Mistral AI's LeChat

This free platform is powered by Mistral Large, and features top-notch Flux image generation and superior web searching – the best, in our opinion, even beating SearchGPT. It supports document/image understanding and open source AI agents, although text quality lags behind competitors. However, Mistral Large LLM is not as powerful as its competitors, making it ideal for power users who want to trade text quality for features.

Honorable Mention: Meta AI, Gemini (from Google Artificial Intelligence Studioand not Main site), Hugging Chat, Rica, Grok-2

Large linguistic models

A large language model, or LLM, is an artificial intelligence system trained on massive amounts of textual data to understand and create human-like language. You can see it as a glorious autocomplete. It is designed to predict the most likely symbol (think words, although that's an imprecise comparison) in the set.

The result is natural text that sounds human because it resembles what humans would do.

Here is our list of the best LLMs so far:

Best Specialist: OpenAI's GPT-4o

It balances creative writing, programming, and reasoning with its customizable “Canvas” feature, despite its style It can feel predictable. The latest release (from November 20) also achieved first place on the list LLM Arena With an ELO score of 1,366, beating the Google Gemini beta released on November 21.

Best for Writing: Anthropic's Claude 3.5 Sonnet

It matches or exceeds GPT-4o in many areas with greater innovation, human-like outputs, Although he is prone to hallucinations.

Best at Storytelling: Long Form Writer

is born 10,000+ word stories Within minutes. Do we need to say more?

Most versatile: Meta's Llama-3.1

the Pioneering open source model With comprehensive customization, LoRA creation and fine-tuning options, available in sizes from 7 billion to 405 billion parameters so users can operate it on their local devices or cloud servers according to their needs. Nvidia has developed a custom version called "Nemotron", which has been generating some buzz in the community and is worth checking out.

Biggest disappointment: Llama-3.1 70B reflection

Announce With high expectations, the model claimed to outperform GPT-4o thanks to the Chain of Thought built into it. It ended up being a huge fiasco with bogus benchmarks, hidden API calls to Claude AI, and... Big controversy.

Image generators

An image generator is basically a form that takes a text input and provides an output associated with that text input. So, for example, you say, “Green horse with dragon face,” and the model will generate an image of a green horse with a dragon face. You can also enter something like "busty waifu", but that's not its purpose.

These are some of the best image generators currently available

Best specialist: Phlox

Flow dominates Latest generation AI models with high customization, LoRA/ControlNet support, and text generation capabilities. It requires powerful hardware, but it shows off a distinct style with intense bokeh and lax surface details that users are still trying to process.

It comes in three flavors: Pro (closed source, the most powerful model), Dev (non-commercial license), and Schnell (open source distilled version). They all offer excellent image generation capabilities, and the ceiling will rise if fine-tuning is taken into consideration.

Best for realism: Recraft v3

Thank you Unparalleled realismoffers versatile presets and better value than proprietary alternatives like MidJourney.

It has a free tier that offers the same quality, even though Recraft has generations.

Best Anime: MidJourney Niji

Unparalleled quality of cartoon style images; Fine-tuning stable diffusion is a secondary option.

Most versatile: stable spread 3.5

Stable diffusion 3.5 Huge improvement Via SD3 with better licensing, detailed output, and additional support.

It's more resource efficient than the Flux for fine-tuning and is a full model – unlike the Flux Schnell, which is a distilled version – making it a better choice for custom models.

However, it came out a bit late and was overshadowed by the popularity of Flux.

Biggest disappointment: SD 3 average

Everyone expected this new model to be the new king of image generators, beating out the SDXL and all the others. He ended up being a bad and notorious role model Awesome license And the horrific deviations when trying to generate them People on the grass.

Video generators

Video generators take the image creation process one step further. They generate each frame and use it as input to generate the next frame with image consistency and high speed commitment.

This is still a work in progress, and models can only create a few seconds of video. Here is a list of some of the best you can try.

Best Specialist: Kling

Rapidly improve the Chinese model, Superior to Sora In some cases. It supports training on facial models, and consistently creates high-quality scenes that show great diversity in terms of styles, realism, and camera movement.

Top contender: Runway Gen 3

Pioneering creative video The app has solid environmental understanding, but struggles with fast-paced scenes.

Best for Storytelling: ShowRunner

We can't tell you much about it this. However, in secret tests he showed enormous potential.

Best Open Source: Genmo Mochi 1

It's great He releases Outperforms competitors like Rhymes Allegro and Stable Video Diffusion with superior realism and frame consistency.

Biggest disappointment: OpenAI Sora

Announce With high expectations as a revolutionary "universal model" that transcends any video generation, it remains disappointingly unavailable today Leaky outputs.

Honorable Mention: Google Veo

Google Show It was released on December 3. We haven't tested it, but the generations Google shares seem pretty nice. Of course, we're on the waiting list to test the model, and you'll be the first to know our thoughts once we have access to it.

Music generators

Just like video generators, music generators create songs. However, it is different from audio generators, since the output is more specialized in melodic output that does not include noise, regular sounds or sound effects.

Users can rely on a separate LLM to generate song lyrics or manually enter the lyrics, set some parameters such as the song style, and then the model outputs the relevant music from scratch.

These two are the best, as well as an open source alternative.

Best Specialist: Suno v4

He excels in vocals and lyrics, variety of style, and long-form consistency. previous one, Sono v3.5But it's not free It remains a strong alternative

Best competitor: Audio

Sono's biggest competitor. It offers amazing composition accuracy, and almost rivals the Suno v4 in vocals. Some generations outperform the Suno v3 In a personal style.

Best Open Source: Stable Audio 2

The open source scene doesn't do much in this area. Static sound 2 It seems to be the best model, but it lags behind closed source competitors across the board. dead audiocraft And MusicGen are alternatives, but far from being industry leaders. The fine-tuners didn't notice, and they're usually the people behind the cherry on top that makes open source models so great.

Modified by Andrew Hayward