ME
Meta Platforms / Meta Superintelligence Labs
Developer and publisher of Muse Code and Muse Spark 1.2; positions the product as a lower-cost alternative to rivals to gain developer adoption, using the contributor tier's data-training rights as an extension of the internal data flywheel Meta already runs.
Competing terminal/agentic coding product; Muse Spark 1.2 edges out GPT-5.6 Terra in Codex on Terminal-Bench 2.1 (82.9% vs 81.8%) but trails on DeepSWE v1.1 long-horizon tasks.
AN
Anthropic (Claude Code)
Market leader Muse Code is explicitly built to challenge; Claude Opus 5 (max effort) leads Terminal-Bench 2.1 at 86.7%, ahead of Muse Code, and Meta's own launch benchmarks show Claude Opus 5 beating Muse Code on every comparison Meta itself published.
XA
xAI (Grok Build / Grok 4.5) and Google (Gemini 3.6 Flash)
Secondary competitors also benchmarked at launch; Grok Build scored 81.6% on Terminal-Bench 2.1, behind Muse Spark 1.2, which also exceeds Grok and Gemini 3.6 Flash on DeepSWE 1.1.
Competing lower-cost model provider (V4 Flash 0731); Meta's launch positions Muse Spark 1.2 as narrowly ahead of DeepSeek V4 Flash on several benchmarks, situating this launch within a broader agentic-coding price war beyond just the Meta/OpenAI/Anthropic rivalry.
Meta CEO; publicly announced Muse Code's beta launch.
Meta Superintelligence Labs AI chief; publicly framed Muse Code's value proposition around cost-effectiveness, saying the tool 'can be an incredibly good option, especially from a cost perspective' for many workflows.