Is Muse Spark Open good for AI agents?
Trying Meta's open-weight coding model through OpenRouter, or self-hosting it where code must stay on your own hardware. On our scoring, which is calibrated to public benchmarks, it rates 5.0/10 for agentic ability and 7.0/10 for output quality. Open weights in BF16 and NVFP4, from Meta.
How much does Muse Spark Open cost?
Muse Spark Open lists at $1.25 per million input tokens and $4.25 per million output tokens on OpenRouter as of October 16, 2026, the same price at Modal, Together, Nebius, Fireworks and Baseten, with cache reads at $0.15 on all but Nebius. Via OpenRouter it is listed at $1.25 in · $4.25 out / 1M tokens.
What is Muse Spark Open's context window?
Muse Spark Open has a 1.0M-token context window, large enough to hold sizeable codebases or document sets in a single run. Meta is the maker.
Is Muse Spark Open open source?
Yes. Muse Spark Open ships with open weights, so you can self-host it or run it through a hosted provider such as OpenRouter. Check the license before commercial use.
What does the Muse Model License allow?
It grants free use, modification, fine-tuning, distillation, redistribution and sale, with conditions. You must keep the license and Meta's Usage Policy with every copy and follow that policy, which bars uses such as malware, weapons and impersonation. Every product or service that uses the model or a derivative must prominently show "Built with Muse Spark Open" in its interface, documentation, website or settings. A company that offers models to others through an API (model as a service) and has more than $20 million in gross revenue over any 12 months must notify Meta and sign a separate agreement first; internal use, research and using the model inside your own product that is not a model API are exempt. Meta claims no ownership of outputs. This is a summary, not legal advice.
How do I use Muse Spark Open in a coding agent today?
The simplest way is OpenRouter, which lists it as meta/muse-spark-open from Modal, Together, Nebius, Fireworks and Baseten. In an agent that takes an OpenAI-compatible endpoint, such as OpenCode, Cline, Kilo Code, Hermes Agent or OpenClaw, set the base URL to https://openrouter.ai/api/v1, use your OpenRouter key and the model ID meta/muse-spark-open. Reasoning is always on there, at medium effort unless you set reasoning_effort; Meta's model card recommends high or xhigh for coding and agentic tasks. Meta also lists Baseten, Fireworks AI, Groq, Modal, Nebius and Together AI as API hosts.
Can I self-host Muse Spark Open?
Yes, with large hardware. Request access to the gated weights on Hugging Face, then serve them with vLLM or SGLang, which expose an OpenAI-compatible API your agent can point at. Meta's model card says it needs at least 700 GB of GPU memory in NVFP4 and 2 TB in BF16, and that it is optimized to run on two Mac Studios with M5 Ultra using MLX. Transformers, TensorRT-LLM, llama.cpp and Unsloth are also supported.
Does Standard Compute serve Muse Spark Open?
Not yet. Standard Compute's models are listed at standardcompute.com/models. On the models it serves, the router sends each step to the cheapest model that can do it, on providers vetted for price, uptime and speed with automatic fallback; opt-in Europe or US rules limit which provider endpoints may serve a key, and prompts are not used for training.
What are the downsides of Muse Spark Open?
The main trade-offs: No independent agent benchmark results yet; the published figures are Meta's own; Self-hosting needs at least 700 GB of GPU memory in NVFP4 (2 TB in BF16), per Meta's model card; The license requires "Built with Muse Spark Open" in your product, and an API business above $20 million in yearly revenue needs a separate agreement with Meta; Not served on Standard Compute yet. It's strongest for trying Meta's open-weight coding model through OpenRouter, or self-hosting it where code must stay on your own hardware.
Best model for OpenCode
Best model for Cline
Best model for Kilo Code
Best model for Hermes
Best model for OpenClaw
