Mistral unveils Forge for custom enterprise LLMs
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Mistral announced the Forge platform at Nvidia GTC, enabling enterprises to train custom LLMs from scratch using their own data.
- Arthur Mensch said Mistral is on track to exceed $1 billion in annual recurring revenue this year, reflecting its enterprise focus.
- Elisa Salamanca explained that Forge lets governments and enterprises customize AI models for specific language, culture, and compliance needs, beyond fine‑tuning or RAG.
- Timothée Lacroix noted that Forge leverages Mistral’s open‑weight model library, including the small Mistral Small 4, allowing customers to prioritize domain‑specific performance.
- Mistral provides forward‑deployed engineers (FDEs) to embed with customers, helping build synthetic data pipelines and evaluate models.
- Mistral has already partnered with Ericsson, the European Space Agency, Reply, Singapore’s DSO and HTX, and ASML to pilot Forge.
Why it matters: Enterprises and governments gain full control over AI behavior and data by training models from scratch, reducing reliance on third‑party providers and addressing language or compliance gaps; the move also pushes Mistral toward its $1 billion‑plus annual recurring revenue target and marks a shift toward in‑sourced AI solutions.
