Meta released Muse Code beta, its first coding agent built on Muse Spark 1.2. The release enters a market led by GitHub Copilot and Amazon CodeWhisperer.
Meta shipped Muse Code beta, its first coding agent, wrapping the Muse Spark 1.2 model. The release puts Meta in direct competition with GitHub Copilot and Amazon CodeWhisperer.
Key facts
- Muse Code is Meta's first coding agent, now in beta
- Built around the Muse Spark 1.2 model
- Competes with GitHub Copilot and Amazon CodeWhisperer
- No pricing or performance benchmarks disclosed
- Meta has not detailed supported IDEs or platforms
Meta has released Muse Code in beta, marking its first dedicated coding agent. The agent system is built around the Muse Spark 1.2 model, adding autonomous code generation and editing capabilities to Meta's AI developer tooling. According to @rohanpaul_ai, the beta launch positions Meta to compete directly with established coding agents from GitHub Copilot and Amazon CodeWhisperer.
The company did not disclose specific performance benchmarks or pricing for Muse Code in the announcement. The beta release focuses on integrating the agent into existing development workflows, though Meta has not detailed which IDEs or platforms are supported initially. The Muse Spark 1.2 model itself was previously known for code generation tasks, but this is the first time Meta has packaged it as a full agent system.
Key Takeaways
- Meta released Muse Code beta, its first coding agent built on Muse Spark 1.2.
- The release enters a market led by GitHub Copilot and Amazon CodeWhisperer.
What This Means for the Developer Tools Market
The coding agent space has seen rapid consolidation, with GitHub Copilot dominating enterprise adoption and Amazon CodeWhisperer pushing into AWS-native environments. Meta's entry adds a third major player with significant model infrastructure and developer community reach. The strategic play appears to be leveraging Muse Spark's existing code capabilities rather than building a new model from scratch, suggesting Meta wants speed to market over technical differentiation.
Meta has not announced whether Muse Code will be free, integrated into Meta's developer ecosystem, or offered as a standalone product. The beta status indicates the company is still testing the agent's reliability and user experience before committing to a broader rollout. Developers and enterprises evaluating coding agents now have another option to consider, though the lack of benchmark data makes direct comparisons difficult.
What to watch
Watch for Meta to release Muse Code performance benchmarks on standard coding agent evaluations like SWE-Bench, and for the company to announce supported IDEs and pricing at its next developer conference. If Meta integrates Muse Code into its open-source Llama ecosystem, expect a faster adoption curve among the open-source developer community.
[Updated 07 Aug via simon_willison]
Meta has disclosed pricing for Muse Spark 1.2, with the standard muse-spark-1.2 model at $1.25 per million input tokens and $4.25 per million output tokens. A discounted muse-spark-1.2-contributor variant, available if users allow Meta to use data for product improvement, costs just $0.10/$0.20 — a steep discount compared to rivals like Gemini 3.6 Flash ($1.50/$7.50) [per Simon Willison]. The model is now available via Vercel's AI Gateway, supporting coding agents like Claude Code and Codex, with no platform fee on inference.
Originally published on gentic.news
Top comments (0)