Mistral AI launches Forge platform to let enterprises build custom AI models from scratch

Mistral bets on 'build-your-own AI' as it takes on OpenAI, Anthropic in the enterprise | TechCrunch

Most enterprise AI projects fail not because companies lack the technology, but because the models they're using don't understand their business. The models are often trained on the internet, rather than decades of internal documents, workflows, and institutional knowledge. That gap is where Mistral, the French AI startup, sees opportunity. On Tuesday, the company announced Mistral Forge, a platform that lets enterprises build custom models trained on their own data. Mistral announced the platform at Nvidia GTC, Nvidia's annual technology conference, which this year is focused heavily on AI and agentic models for enterprise. It's a pointed move for Mistral, a company that has built its business on corporate clients while rivals OpenAI and Anthropic have soared ahead in terms of consumer adoption. CEO Arthur Mensch says Mistral's laser focus on the enterprise is working: the company is on track to surpass $1 billion in annual recurring revenue this year. A big part of doubling down on enterprise is giving companies more control over their data and their AI systems, Mistral says. "What Forge does is it lets enterprises and governments customize AI models for their specific needs," Elisa Salamanca, Mistral's head of product, told TechCrunch. Several companies in the enterprise AI space already claim to offer similar capabilities, but most focus on fine-tuning existing models or layering proprietary data on top through techniques like retrieval augmented generation (RAG). These approaches don't fundamentally retrain models; instead, they adapt or query them at runtime using company data. Mistral, by contrast, says it is enabling companies to train models from scratch. In theory, this could address some of the limitations of more common approaches -- for example, better handling of non-English or highly domain-specific data, and greater control over model behavior. It could also allow companies to train agentic systems using reinforcement learning and reduce reliance on third-party model providers, avoiding risks like model changes or deprecation. Forge customers can build their custom models using Mistral's wide library of open-weight AI models, which includes small models such as the recently introduced Mistral Small 4. According to Mistral co-founder and chief technologist, Timothée Lacroix, Forge can help unlock more value out of its existing models. "The trade-offs that we make when we build smaller models is that they just cannot be as good on every topic as their larger counterparts, and so the ability to customize them lets us pick what we emphasize and what we drop," Lacroix said. Mistral advises on which models and infrastructure to use, but both decisions stay with the customer, Lacroix said. And for teams that need more than guidance, Forge comes with Mistral's team of forward-deployed engineers who embed directly with customers to surface the right data and adapt to their needs -- a model borrowed from the likes of IBM and Palantir. "As a product, Forge already comes with all the tooling and infrastructure so you can generate synthetic data pipelines," Salamanca said. "But understanding how to build the right evals and making sure that you have the right amount of data is something that enterprises usually don't have the right expertise for, and that's what the FDEs bring to the table." Mistral has already made Forge available to partners including Ericsson, the European Space Agency, Italian consulting company Reply, and Singapore's DSO and HTX. Early adopters also include ASML, the Dutch chipmaker that led Mistral's Series C round last September at a €11.7 billion valuation (approximately $13.8 billion at the time). These partnerships are emblematic of what Mistral expects Forge's main use cases to be. According to Mistral's chief revenue officer Marjorie Janiewicz, these include governments who need to tailor models for their language and culture; financial players with high compliance requirements; manufacturers with customization needs; and tech companies that need to tune models to their code base.

Mistral AI makes enterprise push with two new launches

Mistral's new platform lets enterprises build custom models trained on their own data. Mistral AI's newest model in the fully open source 'Small' series attempts to consolidate capabilities of its flagship models. The 'Mistral Small 4' is a hybrid model optimised for a variety of tasks, the French company said, powered with reasoning strengths from the 'Magistral' model, multimodal capabilities from 'Pixtral', and coding from 'Devstral'. The 'Small' models are designed for instruction-taking. The new multimodal model is marketed towards developers wanting to automate coding, explore code bases and code agentic workflows. Enterprises can use 'Small 4' for general chat assistants, document analysis and multimodal analysis, while researchers can conduct mathematics, research and other complex reasoning tasks. Small 4 has a total of 119bn parameters with 6bn active ones per token. It automatically switches between models depending on the task, leading to a 40pc reduction in end-to-end completion time and three-times more requests per second when compared to Mistral Small 3. Mistral Small 4 with reasoning matches or surpasses OpenAI's GPT-OSS 120B on long context reasoning, live coding and mathematics benchmarks, the company said. The French start-up also launched a new platform called 'Mistral Forge' that lets enterprises build custom models trained on their own data. The new platform is attempting to target the gap between AI models trained on generic data and enterprises who need models for very specific and tailored needs. Mistral has already partnered with the likes of AMSL, Ericsson and the European Space Agency to train models on their proprietary data. Last September, the 2023-founded French AI darling announced a Series C raise of €1.7bn at a €11.7bn post-money valuation led by ASML. Existing investors DST Global, Andreessen Horowitz, Bpifrance, General Catalyst, Index Ventures, Lightspeed and Nvidia took part. Mistral is a founding member of the Nvidia Nemotron Coalition. As part of the initiative, Mistral and Nvidia plan to co-develop frontier open-source AI models. Meanwhile, yesterday (17 March), OpenAI launched GPT-5.4 in Mini and Nano versions. According to the AI giant, GPT‑5.4 Mini "significantly improves" over GPT‑5 Mini across coding, reasoning, multimodal understanding, and tool use, while running more than two-times faster. GPT‑5.4 Nano is the smallest, cheapest version of GPT‑5.4 that optimises to be speed and cost efficient. These follow the launches of GPT-5.4 Thinking, a high-performance, in depth-reasoning model, and GPT-5.3 Instant, for "fast, everyday work". Don't miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic's digest of need-to-know sci-tech news.

Mistral launches Forge for custom enterprise AI models

Mistral has introduced Forge, a new platform designed to enable enterprises to build custom AI models using their own data. The announcement was made during Nvidia's GTC conference, which focuses on AI advancements for enterprise applications. This offering aims to address the common issue where enterprise AI projects fail due to models inadequately understanding specific business contexts. The emergence of Mistral Forge highlights the increasing demand for tailored AI solutions among corporate clients. CEO Arthur Mensch stated that Mistral is on track to exceed $1 billion in annual recurring revenue this year. This focus on customized solutions allows companies to gain greater control over their AI systems and the data on which they are built. Elisa Salamanca, Mistral's head of product, explained that the Forge platform allows enterprises and governments to adapt AI models to meet their unique requirements. Unlike competitors who primarily fine-tune existing models, Mistral enables clients to train models from scratch, potentially enhancing their performance on domain-specific or non-English data. This approach is intended to mitigate reliance on third-party model providers, thereby reducing risks associated with external changes. Mistral Forge allows users to leverage the company's library of open-weight AI models, including the recently released Mistral Small 4. Timothée Lacroix, co-founder and chief technologist, noted that the platform helps clarify the strengths and weaknesses of different model sizes, offering the ability to prioritize specific topics. The platform also comes with the support of Mistral's team of forward-deployed engineers. These engineers collaborate with clients to adapt data and implement models effectively. Salamanca indicated that Forge is equipped with necessary tools and infrastructure to create synthetic data pipelines but emphasized the expertise needed for proper evaluation and data management. Mistral has already partnered with various organizations, including Ericsson, the European Space Agency, and Italian consulting firm Reply. Early adopters of the Forge platform include ASML, the Dutch chipmaker that led Mistral's Series C funding round at an €11.7 billion valuation last year. Mistral anticipates that Forge will cater to diverse applications, such as developing language models for government use, meeting compliance in financial sectors, supporting customized manufacturing processes, and helping tech firms align models with their codebases.

Mistral AI launches privacy-focussed AI tool Forge to target enterprise clients - The Economic Times

Forge is positioned as a more privacy-focussed tool, unlike those from its rivals, including OpenAI, which rely on cloud services. Forge is designed for industries where data privacy is important, such as finance, defence, and manufacturing.Paris-headquartered artificial intelligence (AI) startup Mistral AI made waves at Nvidia's annual GTC 2026 conference with the unveiling of its new platform Forge, which enables companies to build powerful custom AI models using only their own private data, on their own systems. Forge is positioned as a more privacy-focussed tool, unlike those from its rivals, including OpenAI, which rely on cloud services. Forge is designed for industries where data privacy is important, such as finance, defence, and manufacturing. It simplifies the creation of tailored AI by starting with Mistral's proven open models and customising them to fit specific business needs, such as handling multiple languages or industry-specific tasks. For Mistral, the new tool marks a shift from sharing AI models to powering enterprise AI strategies. CEO Arthur Mensch said the company is on track to cross $1 billion in annual revenue by 2026, moving beyond free models to high-value enterprise deals. Early users of the tool include chip equipment maker ASML (one of Mistral's backers), telecom leader Ericsson, the European Space Agency, Singapore's defence agencies, and Italian consultancy Reply. The launch comes as Nvidia pushes enterprises to adopt its powerful new GPUs. Mistral positions itself as the top choice for companies wanting customisable AI that runs on Nvidia hardware without relying on US cloud giants. This will drive demand for Nvidia GPUs for on-premises training. The French company raised over $2 billion in Series C funding in September 2025, taking its valuation to $14 billion and making it the most valuable European AI company. Mensch had told ET in an interaction earlier this year that Mistral is looking to partner with sovereign cloud providers in India.

Mistral AI Now Lets Enterprises Build Their Own Custom Language Models

This appears to be a well-thought out move as experts have often blamed lack of business understanding as the reason for AI project failure French AI startup Mistral AI announced a new platform that would allow enterprises to build their own custom language models trained on their own data. Called the Mistral Forge, the company is hoping that it would bridge the gap between models trained on internet data instead of documents, workflows and institutional knowledge of an enterprise. The company made this announcement at the Nvidia GTC event in California where several similar reveals have been made related to the field of AI, including the GPU-makers themselves coming out with their own version of OpenClaw as well as announcing a new chip featuring the power of Groq. For a company that is building its revenue models around corporate clients, like competitor Anthropic has done thus far and OpenAI is now treading the same path, the move appears to be a well-thought out one as in the recent past, experts have often ascribed the failure of AI projects to the lack of business understanding among the foundational models. Mistral AI CEO Arthur Mensch believes that his company's focus on the enterprise AI is working well and could lead them beyond the $1 billion annual recurring revenues mark this year. He also noted that a big part of this doubling down on enterprise has also given enterprises more control over their data and AI systems. Elisa Salamnca, Mistral's head of product, was quoted by TechCrunch as saying that what Mistral Forge does is to let enterprises and governments customise their AI models for their very specific requirements. Of course, the French AI startup isn't the only one offering such capabilities as several others focus on fine-tuning existing models or building on top of it via techniques like the retrieval augmented generation or RAG. This option doesn't retrain models but adapt or query them at runtime with company data - a bit like learning on the job. What Mistral is attempting to do differently is allow enterprises to train models from scratch and thereby addressing limitations of the current approaches that include handling non-English or even highly domain-specific data to get better control over model behaviour. The startup believes that this would also help companies train agentic systems with reinforcement learning and cut down reliance on third-party model providers. According to the company, customers would be able to build custom models with Mistral's library of open-weight AI models that includes their smallest one called Mistral Small 4. This would allow companies using Mistral Forge to get more value out of their existing models as well. The idea behind smaller models is that they aren't good across multiple topics as the larger ones, but they are more easy to customise, which means that the human in this chain can direct them to pick what are must-haves, good-to-haves and can be ignored. The company is also offering consultancy services on which models and infrastructure to use, though they are finally decisions taken by the customer. The company said Mistral Forge is already available to partners that include the European Space Agency and Dutch chipmaker ASML that led Mistral's Series C round last year at a $13.8 billion dollar valuation. A few others also have access to the platform such as Ericsson, Singapore companies DSO and HTX and Italian consulting form Reply. By the look of it, the move appears to be yet another well-thought out one as they could well reflect the areas where Mistral Forge could find invaluable use cases for further expansion. So, we have a government agency, a couple of financial companies with high compliance needs and a manufacturer with bespoke requirements.