ChatGPT
OpenAIThe best-known general-purpose AI assistant: writing, analysis, code, images and web search in a single interface.
Compare · Test · Choose
Detailed sheets, independent comparisons, reviews, prompt examples and hands-on tests — for every tool on the market and every profession.
The closed large language models: available through an interface or an API, but whose weights remain the property of their publisher.
The best-known general-purpose AI assistant: writing, analysis, code, images and web search in a single interface.
Google's AI built into Workspace: very long context window, native multimodality and deep ties to Google services.
Anthropic's assistant, renowned for its writing quality, reasoning and coding abilities (Claude Code).
Microsoft's AI everywhere: in Windows, Office 365 and the code editor with GitHub Copilot, the developers' assistant.
xAI's assistant with real-time access to X: irreverent tone, strong reasoning and live knowledge of what's happening.
Models whose weights are published: downloadable, auditable, fine-tunable on your data and hostable on your own infrastructure.
The French champion: strong open models, data hosted in Europe and a highly competitive API.
Moonshot AI's assistant: open-weight K2 models with agentic skills and very long documents handled with ease.
The lab that shook the market: frontier-level open-weight models (MIT license) at a fraction of the usual cost.
Alibaba's open-weight model family: every size from mobile to server, Apache licensed, a favorite of the open source community.
Meta's open-weight family that started the movement: free to download, fine-tune and deploy on your own hardware.
The latest announced models, currently being tested by our editorial team — their full sheets are coming soon.
OpenAI
The latest generation of OpenAI's flagship model: stronger reasoning and long-term memory.
The preview of Google's next pro model: native multimodality and an extended context window.
Alibaba Cloud
Alibaba's open-weight model, a community-benchmark reference, deployable from mobile to server.
Mistral AI
Mistral AI's frontier model: top-tier performance with European hosting.
Anthropic
The first model of Anthropic's Claude 5 family, in the new Mythos class above Opus in capability.
Google's video generation model: realistic clips with synchronized sound from a text prompt.
Google's viral image editing model: consistent characters and precise edits in natural language.
OpenAI
OpenAI's next image model: sharper text rendering and finer instruction following.
Beyond LLMs: image generation, design, voice and audio — the closed tools that lead their category.
The reference for image generation: unmatched artistic rendering, mastered styles and a massive creative community.
Adobe's generative AI, trained on licensed content and built into Photoshop: the commercially safe choice for design teams.
The reference for AI voice: ultra-realistic text-to-speech, voice cloning and dubbing in dozens of languages.
ChatGPT vs Claude, Gemini vs Copilot… detailed head-to-heads to choose with confidence.
Ready-to-copy prompts, organized by tool, profession and difficulty level.
Our protocols applied to the market's tools, with dated, reproducible results.
Lawyer, marketer, developer, designer… the best tools and use cases for your job.
Proprietary and open-weight large language models: the ones we tested and the ones coming up.
Open-weight, auditable, self-hostable models — to keep control of your data.