Meta just put a new coding assistant in the ring. Muse Code, launched on August 6 and currently in beta, runs on Meta’s Muse Spark 1.2 model and targets developers building and debugging software.
What It Does
Meta says Muse Code can write code, verify results, and handle long complex projects. It runs multiple sub-agents simultaneously to work through difficult tasks faster. The tool also keeps a running log of its actions, so if it crashes mid-session it picks up where it left off rather than starting from scratch.
Meta trained Muse Spark 1.2 and Muse Code together as a paired system, with the goal of smooth operation between the model and the assistant layer.
️ Pricing
- Standard tier: $1.25 per million input tokens
- Output: $4.25 per million output tokens
- Model: pay-as-you-go, no flat subscription mentioned
Context
Muse Spark 1.1 shipped roughly a month earlier for developer testing. Meta used that earlier model to generate and evaluate difficult coding challenges, and the results fed directly into improving Muse Spark 1.2’s ability to follow complex instructions. Muse Code competes in the same space as Anthropic’s Claude and OpenAI’s Codex.
