Mind Lab Launches Mint Recursive Platform for Enterprise AI Models

SAN FRANCISCO, Oct. 21, 2026 /CourierPR/ -- Mind Lab, an AI research lab focused on post-training and inference, has launched a new platform, Mint Recursive, to assist enterprises in building industry-specific AI models. The platform, which covers benchmark design, data, post-training, inference, and deployment in a single workflow, is now available to companies across various sectors, including pharmaceuticals, finance, the public sector, and gaming.

The platform begins each engagement by building a benchmark with the customer, defining the task goals, correct behavior, common mistakes, and the acceptance criteria a model must meet. Data development, LLM post-training, and deployment then run in a single loop, ensuring each iteration is measured against a standard set by the customer. This approach aims to lower costs and streamline the process for teams that might otherwise struggle to train their own models.

Mind Lab combines LoRA post-training, shared base models, and distributed optimization to reduce the number of parameters that need to be updated and the amount of compute required. According to the company, GPU costs on shipped projects have fallen by more than 90%, and specialist training starts at just $1,000. Customers can choose from a range of base models, including GLM, DeepSeek, Kimi, Qwen, and MiniMax, and deploy in private, VPC, or hybrid environments.

The company's platform is designed to support a variety of industries, with particular emphasis on healthcare and life sciences. In this sector, Mind Lab assists with tasks such as literature retrieval, evidence review, drug discovery, and R&D intelligence, as well as medical chart quality control. The platform has been used to cover 27 quality-control indicators and more than 8,000 physician-experience traces, significantly improving first-attempt delivery success rates.

Mind Lab's founder, Chen Kaijie, emphasized the importance of incorporating a company's expert experience into custom AI models.

The platform grew out of the lab's work in reinforcement learning. When Mind Lab introduced Mint Recursive in September, it also released Macaron-V1.1, a 752B-parameter agent model post-trained from GLM-5.3, with open weights. Moving from Macaron-V1 to Macaron-V1.1 took just two weeks on the platform, and across 60 UI tasks, first-attempt delivery success rose from 71.67% for GLM-5.3 to 96.67% for Macaron-V1.1.

Companies interested in building proprietary AI models can start with a single real task and contact Mind Lab to define the benchmark for it. The platform offers flexibility, allowing customers to choose how much of their own team to commit. Mind Lab's forward-deployed engineers (FDEs) can run the benchmark, data, post-training, and deployment end to end, while teams that define their own model, data, and configuration can hand scheduling, training, and delivery to the platform's automated training. Teams that want full control can write their own loss functions and training loops, while the lab provides clusters, trajectory sampling, weight sync, and inference.

Mind Lab is an AI research lab focused on LLM post-training, with a research center on experiential intelligence. The lab developed Mint Recursive, a platform for post-training and inference, and also developed the Macaron series of agent models, releasing Macaron model weights as open source on Hugging Face.