MuseNet vs. Muse Spark 1.2

We have introduced MuseNet, a sophisticated deep neural network that can generate 4-minute compositions using ten unique instruments, effortlessly integrating genres from country music to the timeless works of Mozart and even the legendary tunes of the Beatles. Instead of being explicitly programmed with musical principles, MuseNet discerns and internalizes patterns of harmony, rhythm, and stylistic nuances by predicting the next note in an extensive database of MIDI files. This cutting-edge model utilizes the same unsupervised learning techniques as GPT-2, a powerful transformer model aimed at forecasting the subsequent element in a sequence, applicable to both audio and text. With MuseNet's ability to grasp various musical styles, we can produce distinctive combinations of musical creations. We look forward to seeing how musicians, as well as individuals without formal training, will creatively utilize MuseNet to generate original works! Users have the option to choose a particular composer or style, and they may start with a familiar piece, enabling them to explore the diverse spectrum of musical styles that the model can generate. This not only enhances artistic creativity but also provides a platform for innovative experimentation in the world of music. The versatility and adaptability of MuseNet promise to inspire countless new musical adventures.

Muse Spark 1.2 is a coding-focused AI model from Meta designed to support advanced software engineering tasks through Muse Code and the Meta Model API. The model builds on Muse Spark 1.1 with improvements in code generation, complex debugging, codebase understanding, and end-to-end developer workflows. Muse Spark 1.2 powers Muse Code, a terminal coding agent that can plan repository changes, write code, validate outputs, and work across large codebases. Muse Code uses persistent async background agents that stay active throughout a session to reduce redundant information gathering and support difficult multi-step work. The runtime uses a local event log where model calls, tool runs, approvals, and edits are appended, making sessions replay-exact and restart-safe. Muse Spark 1.2 was co-trained with Muse Code so the model can take advantage of its toolset, harness workflows, goals, compaction, and subagent architecture. Meta significantly scaled training compute on coding tasks and expanded training environment diversity to improve the model’s engineering capabilities. The model was also trained on long-horizon coding tasks, including whole-repository generation, large end-to-end projects, auto-research, and extended iterative work. Its training approach uses planning, goal conditioning, context compaction, rejection-sampled harness trajectories, and self-improvement data generated with Muse Spark 1.1. Meta also tested Muse Spark 1.2 on long-running GPU kernel optimization workflows where the model wrote, compiled, profiled, and improved Triton kernels over many tool calls. By combining coding-focused training, agentic runtime integration, persistent subagents, long-horizon reasoning, replay-safe execution, and API availability, Muse Spark 1.2 helps developers and AI agents complete complex software engineering work with less intervention.