Labrador CMS adds Mistral AI, and new models
GPT-5.4 becomes the new default as Labrador expands its AI model lineup
Published
Modified
Labrador CMS is adding native support for Mistral, the Paris-based AI company that has emerged as Europe's leading large language model provider. With this release, Labrador supports ten models across four AI providers: OpenAI, Anthropic, Google, and Mistral.
GPT-5.4 is also being set as the new default model for all installations.
"We're seeing increasing demand for alternative models, for a range of reasons. We already support ChatGPT, Gemini, and Claude. It was time to natvely add the best European model as well," says Jan Thoresen, CEO of Labrador CMS.
The full model lineup
OpenAI — ChatGPT
GPT-5.4 — Advanced, Multimodal, Improved (new default)
GPT-5.4 Mini — Lightweight, Efficient, Quick
GPT-5.4 Nano — Ultra-Fast, Cost-Effective, Compact
Anthropic
Claude Sonnet 4.6 — Balanced Performance, Efficient, Intelligent
Claude Haiku 4.5 — Fastest, Cost-Effective, Compact
Google — Vertex AI
Gemini 2.5 Pro — Most Capable, Reasoning, Multimodal
Gemini 2.5 Flash — Balanced, Fast, Efficient
Mistral AI (new)
Mistral Large — Powerful, Flagship, Advanced
Mistral Medium — Balanced, Versatile, Efficient
Mistral Small — Fast, Cost-Effective, Compact
Who is Mistral?
Mistral AI was founded in Paris in April 2023 by three researchers — Arthur Mensch, Guillaume Lample, and Timothée Lacroix — who previously worked at Google DeepMind and Meta. The company is now valued at more than $14 billion and has grown into one of the most prominent AI labs outside the United States.
"We're seeing increasing demand for alternative models, for a range of reasons. We already support ChatGPT, Gemini, and Claude. It was time to natvely add the best European model as well,"
What sets Mistral apart is its commitment to open-weight models and European infrastructure. European startups and public institutions that face regulatory or data sovereignty constraints often cannot use services like OpenAI — Mistral has positioned itself as the credible alternative. Mistral Large 3 was trained on a wide variety of languages, with particular strength in European ones, making it a natural fit for publishers serving multilingual audiences. In March 2026, the company raised $830 million to build new datacenters near Paris and in Sweden.
What this means for newsrooms
All of Labrador's AI features — text assistance, translation, article summarization, AI-generated image metadata — work across all ten models. Publishers choose the model that fits their infrastructure requirements, cost profile, or editorial preferences.
For newsrooms operating under data residency requirements or procurement policies that favor non-US vendors, Mistral opens options that weren't available before.
Model selection is configured in the admin settings. All models work with Labrador's prompt customization layer, where publishers define their own tone of voice and brand guidelines — so generated output sounds like the publication regardless of which model is running.
GPT-5.4 replaces the previous default. Publishers can verify or change their model selection at any time in the AI settings panel.
Labrador CMS serves 350+ publications in 20+ countries. All AI features are included in the standard subscription up to a certain level. Though, we strongly urge customers to add their own API keys to maintain data integrity.
