AI Model Timeline

Tracking the accelerating release frequency of frontier AI models.

Muse Code

Meta·

LLMsspecializedcloud

Overview

Meta’s first coding agent: a terminal-based harness for macOS and Linux, released in beta and powered by Muse Spark 1.2. Runs a simple agent loop plus persistent async background agents that stay alive for a whole session and accumulate context instead of restarting per task; larger tasks are split onto parallel subagents in isolated git worktrees so the working copy stays untouched. Meta reports one run optimising Nvidia Hopper GPU kernels across more than 1.000 tool calls in sessions of up to 24 hours. Pay-as-you-go at Muse Spark 1.1 API rates, or a contributor tier at over 90% discount in exchange for training on prompts and completions.

Capabilities and innovations

Terminal coding agentRepository-scale planning, editing and validationPersistent async background agentsParallel subagents in isolated worktreesLong-running sessions (24h+)Session-persistent background agentsContributor tier (data-sharing discount)

Benchmarks

No benchmark values are recorded for this model.

Links

More from Meta

Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.