Malted AI

Malted AI develops specialised small language models (SLMs) for enterprises, using knowledge distillation to transfer capability from large language models into smaller, domain-focused models. The company positions distilled models as an alternative to general-purpose foundation models, arguing that a single many-billion-parameter model is costly, overly general and slow for complex enterprise tasks; its models are described as targeting a single task with high accuracy at 10-100x cost savings versus large models [4][6][7]. The company's platform enables enterprises to build custom AI solutions using distilled LLMs, injecting domain-specific knowledge into smaller models that can run in production [2][5].

The company's current commercial focus is financial services. Its platform ingests customer interactions across calls, chats and emails and analyses them in real time, in place of retrospective sampling, so that institutions can oversee customer outcomes and evidence compliance with regulatory expectations such as Consumer Duty. Deployment options include dedicated single-tenant environments with data residency choices [0][1]. Early pilots were run with financial and legal firms, with wider rollout planned through 2024 [6][7].