We introduce a direct model for speech recognition that assumes an unstructured, i.e., ﬂat text output. The ﬂat model allows us to model arbitrary attributes and dependencies of the output. This is different from the HMMs typically used for speech recognition. This conventional modeling approach is based on sequential data and makes rigid assumptions on the dependencies. HMMs have proven to be convenient and appropriate for large vocabulary continuous speech recognition. Our task under consideration, however, is the Windows Live Search for Mobile (WLS4M) task . This is a cellphone application that allows users to interact with web-based information portals. In particular, the set of valid outputs can be considered discrete and ﬁnite (although probably large, i.e., unseen events are an issue). Hence, a ﬂat direct model lends itself to this task, making the adding of different knowledge sources and dependencies straightforward and cheap. Using e.g. HMM posterior, m-gram, and spotter features, signiﬁcant improvements over the conventional HMM system were observed.