The organisation also announced it is actively expanding the robotics team and is looking to recruit talented research scientists and engineers.
France’s Mistral AI has announced the launch of a new robotics navigation model, as the company further expands in the physical AI space, following deals with a number of key players in Europe’s industrial and manufacturing sector, such as Airbus SE and BMW.
The new 8B model, Robostral Navigate, allows robots to autonomously move around in complex environments via a single RGB camera and basic language prompts. Combining pointing-based navigation with continuous learning elements, the hardware is also agnostic meaning it can be deployed across any robotics fleet.
Mistral claims that the model, prompted by a single instruction, can complete the entire task on its own, moving through a live space full of people and obstacles it was never shown, adapting to any setting. Spaces in which it can be used includes offices, residential and commercial buildings and outdoor settings.
In a post announcing the launch, Mistral said, “We leverage our knowledge of post-training LLMs at scale, using online reinforcement learning, to boost the performance of Robostral Navigate. After the supervised training stage, we further improve the model’s performance using CISPO, an online reinforcement learning algorithm.
“This enables the model to learn from trial and error, recover from failures, and acquire exploratory behaviours, effectively mitigating the distribution shift issue of vanilla behaviour cloning. This alone improved the success rate by 3.2pc. We are not seeing any plateauing, so we are confident that more training and more experiments will continue to push this number up.”
A leader in Europe’s AI space, Mistral is also positioned as a key rival for US counterparts, such as Anthropic and OpenAI. In March of this year, the company raised $830m in its first debt financing, with the intention of funding a new data centre near Paris.
It was announced that the deal, which was supported by a consortium of seven global banks, would pay for Nvidia Grace Blackwell infrastructure with 13,800 Nvidia GB300 GPUs at the “cutting-edge” centre, bringing powered capacity to 44MW.
Previously commenting, Arthur Mensch, the CEO of Mistral AI, said, “Scaling our infrastructure in Europe is critical to empower our customers and to ensure AI innovation and autonomy remain at the heart of Europe.”
Mistral is also looking to recruit, with plans to expand the robotics team. Currently it is aiming to hire additional research scientists and engineers.
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.
You must be logged in to post a comment Login