spb/tools-spboucher-ai

1{2  "slug": "mistral-ai",3  "name": "Mistral AI",4  "aliases": ["La Plateforme", "Mistral Studio", "Le Chat", "Vibe"],5  "vendor": { "name": "Mistral AI", "slug": "mistral", "website": "https://mistral.ai" },6  "homepage_url": "https://mistral.ai/",7  "docs_url": "https://docs.mistral.ai/",8  "repo_url": "https://github.com/mistralai/client-python",9  "pricing_url": "https://mistral.ai/pricing/",10  "product_type": "platform",11  "product_status": "active",12  "summary": "French model provider offering open-weight and proprietary models through the Studio API, OCR and speech models, the Vibe agent for productivity and coding, and self-deployment options",13  "description": "Mistral AI is a Paris-based model developer whose platform is organized into Studio (the developer console and API with playground, evaluations, agents and SDKs), Vibe (a unified agent for productivity and coding on web, mobile, terminal and editor) and Admin (organization, billing and SSO controls). The model catalog mixes its own releases such as Mistral Large and Mistral Small, Devstral for coding, Voxtral for transcription and an OCR model with document layout output, with third-party open-weight models hosted under their own licenses. Several Mistral models are published as open weights, some under Apache-2.0 and newer large models under a Mistral-specific open-weight license that distinguishes research from commercial use.\n\nThe API bills per token with batch and cached-input discounts, OCR per page and speech per minute; consumer and team plans for Vibe run Free, Pro, Team and Enterprise, with an education discount. Enterprise customers can deploy in their own cloud or on-premises, and the open-weight models run on engines such as vLLM or llama.cpp. Mistral is attractive where European hosting, open weights or a hybrid of hosted and self-run models matter; its catalog is smaller than the largest US labs and license terms differ per model, so each should be checked.",14  "capabilities": ["Studio API with chat, embeddings, OCR, transcription, agents and fine-tuning", "Open-weight models deployable on your own hardware", "Vibe agent for documents and coding across web, mobile, CLI and editor", "Batch processing and cached-input pricing", "Admin console with SSO, workspaces and access policies", "Enterprise private cloud or on-premises deployment"],15  "workflows": ["Run an open-weight Mistral model locally and switch to the API for larger ones", "Extract text and layout from scanned PDFs with the OCR model", "Use Vibe in the terminal for coding tasks"],16  "advantages": ["Open-weight options alongside the hosted API", "European vendor with EU hosting", "Document OCR and speech models in the same platform"],17  "limitations": ["Licenses vary by model; newer large models restrict commercial self-deployment", "Smaller third-party ecosystem than OpenAI or Anthropic", "Rapid product renames (La Plateforme to Studio, Le Chat to Vibe) date older docs"],18  "prerequisites": ["Mistral account and API key, or hardware to run open weights"],19  "audiences": ["European organizations with hosting constraints", "Developers wanting open weights with a hosted fallback", "Teams needing OCR and speech alongside text models"],20  "alternatives": ["openai-platform", "claude-developer-platform", "gemini-api", "together-ai", "ollama"],21  "categories": [{ "slug": "model-providers", "primary": true }, { "slug": "ai-assistants", "label": "Vibe (formerly Le Chat)" }],22  "tags": ["llm", "api", "open-weights", "ocr", "speech", "python", "typescript"],23  "use_cases": ["build-agents", "run-local-models", "write-code-with-ai"],24  "operating_systems": ["any"],25  "execution_modes": ["cloud", "self-hosted"],26  "interfaces": ["api", "sdk", "web", "cli", "mobile"],27  "deployment_models": ["saas", "on-premises", "managed-cloud"],28  "integrations": ["vLLM", "llama.cpp", "Ollama", "Hugging Face Hub", "LangChain", "LiteLLM"],29  "data_formats": ["JSON", "PDF (OCR)", "Audio", "Images"],30  "languages": ["Python", "TypeScript"],31  "source_model": "mixed",32  "pricing_model": ["freemium", "usage-based", "paid", "enterprise"],33  "pricing_checked_at": "2026-11-02",34  "self_hostable": true,35  "sources": [36    { "url": "https://mistral.ai/", "type": "official_site", "title": "Mistral AI", "supports": ["name", "summary"], "checked_at": "2026-11-02" },37    { "url": "https://docs.mistral.ai/", "type": "official_docs", "title": "Mistral AI documentation", "supports": ["capabilities", "interfaces", "source_model"], "checked_at": "2026-11-02", "note": "Describes Studio, Vibe and Admin; lists per-model licenses" },38    { "url": "https://mistral.ai/pricing/", "type": "pricing_page", "supports": ["pricing_model", "deployment_models"], "checked_at": "2026-11-02" },39    { "url": "https://github.com/mistralai/client-python", "type": "repository", "title": "client-python (Apache-2.0)", "supports": ["languages"], "checked_at": "2026-11-02" }40  ],41  "confidence": "high",42  "uncertainty_note": "No single license_spdx applies: the hosted platform is proprietary and open-weight models carry per-model licenses (Apache-2.0 for some, a Mistral open-weight license for others).",43  "discovery_method": "editorial-research",44  "depth": "standard",45  "editorial_status": "reviewed",46  "visibility": "public",47  "featured": false,48  "author": "batch-A"49}50