Paris Soir - Generative AI rivals racing to the future

GenAI tools are able to create images, videos, or written works as well as answer questions or tend to online tasks based on simple prompts.

These AI assistants stand out for their popularity and sophistication.

- Hot ChatGPT -

AI existed before ChatGPT, but it was first to make GenAI freely available for people to use as a dedicated application.

San Francisco-based OpenAI has made ChatGPT more powerful and capable with each update, the most recent being GPT 4.5.

One version of ChatGPT released late last year, called o1, was touted as a new-generation that takes time to ponder answers, providing comprehensive results and less inclined to err.

Instead of instantly cranking out results, the model shares its "chain of thought".

OpenAI has imbued ChatGPT with the ability to act as a digital "agent" capable of browsing the internet, compiling information and using computers the way people do when working on tasks.

- Google Gemini -

Google has long put AI to work behind the scenes at its platform but cranked out Bard to take on ChatGPT in March of 2023.

Bard was gradually replaced by a more advanced Gemini model built into Pixel phones and more.

The Internet giant integrated Gemini into its famous search engine to display results summaries called "AI Overviews" along with links in response to queries.

Google also put AI to work letting people search using pictures, video, or sound instead of just typed words.

Such "multimodal" input capability has become common in GenAI tools.

A Gemini 2.0 model capable of "step-by-step" reasoning made its debut in February of this year.

- Cautious Claude -

Founded by former OpenAI engineers, Anthropic launched Claude in March 2023.

The San Francisco-based startup stresses responsible development of AI, moving more cautiously than competitors as it innovates.

Anthropic unveiled Claude 3.7 Sonnet in February, its first model combining instant responses and thoughtful reasoning.

Claude was previously enhanced with a "computer use" feature that let the AI independently perform computer tasks as a person might.

- Mighty Meta -

Meta has integrated custom AI into Facebook, Instagram, Threads, WhatsApp, Messenger and its Ray-Ban connected glasses with the aim of making it the most widely used digital assistant in the world.

Meta's chatbot is based on the tech firm's open-source Llama model, considered one of the most powerful in the world.

Recent press reports tell of plans by the Silicon Valley titan to release MetaAI as a stand-alone application in a direct challenge to OpenAI and Google.

- Grok Snark -

A co-founder of OpenAI, Elon Musk cut ties with the startup in 2018. Since ChatGPT took the lead in the GenAI race, Musk has sued OpenAI, offered to buy it, and launched a rival named xAI.

Musk's chatbot Grok has the advantage of being able to use the trove of posts at X, formerly Twitter, for training the AI model.

The Tesla tycoon bought Twitter in late 2022.

Musk made up for lost time by spending billions of dollars on high-end Nvidia chips for powering AI datacenters.

He promotes Grok as a chatbot with personality, humor and fewer constraints on what it produces.

- Upstart DeepSeek -

DeepSeek was founded in 2023 by Chinese investment fund High-Flyer. In January 2025, the Hangzhou-based start-up turned the world of generative AI upside down with its R1 model.

DeepSeek claims the AI tool was built using less sophisticated chips than its competitors, slashing the cost.

The application was downloaded tens of millions of times in just a few weeks.

- Mounting mix -

Chinese tech behemoths Tencent (Yuanbao), Baidu (Ernie) and ByteDance (Doubao) are also vying for position in the AI market. In early March, Alibaba released its QwQ-32B model, which it claims matches the performance of DeepSeek-R1.

France-based Mistral early last year released Le Chat, AI software particularly advanced in document and image analysis.

R.Poirier--PS