Muse Spark 1.3 aims to compete with Opus and GPT 5.6 Sol
Meta introduced Muse Spark 1.3, a model designed for agentic workflows and optimized for competitive coding performance. The model offers higher first-attempt accuracy, reliable tool calling, and native multimodal perception of video, images, and documents.
Interesting to see how Meta is trying to compete with top AI models. However, it's a shame that most people won't try them on their own hardware, since memory requirements have made many open source models simply inaccessible. Too bad.
What is an agentic workflow and why does it matter?
An agentic workflow enables AI models to work autonomously on extended tasks, maintain context, handle conflicting inputs, and adapt to changing conditions. It's important for applications where AI functions as an independent assistant without constant human guidance.
Who are the primary target users for Muse Spark 1.3?
The model targets developers interested in multi-agent programming, autonomous coding agents, and AI as a development partner. It focuses on competitive programming and real-world software engineering.
How can developers access Muse Spark 1.3?
The model is available through Meta Model API with OpenAI SDK-compatible clients, through Muse Code for terminal use, and via OpenRouter. It offers two pricing options depending on whether Meta can use the data to improve products.
- Meta AI releases Muse Code and Muse Spark 1.2 for advanced coding tasks — research.meta.ai 81 % match
- Meta releases Muse Glimmer, a small open-weight AI model for local agents — research.meta.ai 76 % match
- Google introduces Gemini 3.7 Flash — blog.google 71 % match
- Muse Spark 1.3
- Meta8
- Muse Code
- Meta Model API
- OpenRouter5
- OpenAI26
- GitHub15