Meta Debuts Muse Code, Its First AI Coding Agent to Take on Anthropic and OpenAI

My take: Meta has entered the AI coding agent market with Muse Code, and that shifts the landscape for any team already using Claude Code, Codex, or another AI programming assistant. More serious competitors mean better pricing and more options, and that benefits everyone.

In benchmarks published by Meta, Muse Spark 1.2 scored 82.9% on Terminal-Bench 2.1, behind Claude Code at 86.7%, but ahead of OpenAI's Codex at 81.8% and xAI's Grok Build at 81.6%. That said, this data comes from Meta's own documentation, so it is worth waiting for independent evaluations before using these numbers as a definitive reference.

The pricing strategy is the most interesting part. The standard tier costs $1.25 input and $4.25 output per million tokens. There is a "Contributor" tier 21 times cheaper on output ($0.20), with one concrete condition: Meta uses your code traffic to train future models. For projects with proprietary or confidential code, that trade-off deserves careful thought.

Is your team already using an AI coding agent in its daily workflow? If not, this is a good time to compare options.