The ecosystem counts hundreds of tools. Rather than listing them all, this module organises them by use, pointing out the best-known players.
Two major distinctions structure the market: proprietary vs open-weight, and general-purpose vs specialised (vertical).
3.1Conversational assistants and writing (text)
The Swiss Army knives of generative AI: ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Le Chat (Mistral, French), Copilot (Microsoft), Grok (xAI). Uses: writing, summarising, translation, brainstorming, document analysis.
Advantages: versatility, considerable time savings. Limits: hallucinations, dependence, data sent to third-party servers.
3.2Image generation and editing
Midjourney (artistic quality), ChatGPT Images / DALL·E (OpenAI, strong at text within images), Nano Banana (Google, photorealism), Stable Diffusion and Flux (Black Forest Labs, open-weight, can run locally), Ideogram, Recraft, Reve.
Uses: illustration, design, visual prototyping, advertising. Limits: copyright and the provenance of training data.
3.3Video generation
Veo (Google, the leader in 2026), Runway, Kling, Luma, Pika. (Note: OpenAI's Sora was withdrawn in April 2026.) Uses: marketing clips, storyboards, special effects. Limits: temporal consistency, cost, risk of deepfakes.
3.4Voice, audio and music
ElevenLabs (the reference in speech synthesis), Suno and Udio (music generation), Cartesia, Voxtral (Mistral). Transcription: Whisper (OpenAI), Deepgram, AssemblyAI. Uses: dubbing, podcasts, audiobooks, accessibility.
3.5Programming and development
The area where AI has the most measurable impact: GitHub Copilot, Cursor, Claude Code (Anthropic), Codex (OpenAI), Windsurf, as well as platforms that build applications from a simple description (Lovable, v0, Bolt, Replit).
Advantages: higher productivity, access to code for non-developers. Limits: technical debt, security flaws, over-confidence in code nobody reviewed.
3.6Research and productivity
Perplexity (a sourced conversational search engine), NotebookLM (Google, summaries of your own documents), Deep Research features built into the major assistants, office copilots (Microsoft 365 Copilot, Google Workspace).
3.7Autonomous agents
The new frontier (see Module 7): systems that do not just answer but act, browse the web, fill in forms, operate a computer ("computer use"). Examples: agents built into the major models, Manus, and multi-agent architectures in companies.
3.8Vertical tools (specialised by sector)
- Legal: Harvey, Spellbook.
- Health: Nabla (French, consultation assistant), Abridge.
- Marketing / sales: Jasper, Copy.ai.
- Finance, HR, customer support: dozens of specialised SaaS products.
3.9Platforms for developers
Hugging Face (the "public square" of open-weight models), the APIs of the major labs, and aggregators such as OpenRouter, which give access to many models through a single interface.
How to choose?
Three questions guide the choice: Which modality? (text, image, code…); What level of privacy? (an open-weight model run locally sends no data out); General-purpose or specialised? (a vertical tool is often more effective and better integrated for a specific task).
Sources
New AI Model Releases, June 2026 (blog.mean.ceo); release tracking on llm-stats.com; official vendor documentation.
Finished this module?
Your progress is saved on this device.