{"schema":"postcutoff/model@1","as_of":"2026-10-15T23:34:00+02:00","url":"https://postcutoff.com/m/muse-spark-open/","md":"https://postcutoff.com/m/muse-spark-open/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"muse-spark-open","name":"Muse Spark Open (816B-A42B)","org":"Meta","family":"Muse","released":"2026-10-15","status":"current","type":"reasoning-llm","modality_in":["text","image"],"modality_out":["text"],"open_weights":true,"model_license":"muse-model-license (limited commercial terms)","context_window":1000000,"max_output":null,"knowledge_cutoff":"2026-01","pricing":{"input":1.25,"output":4.25,"unit":"per 1M tokens (USD) on OpenRouter at launch","source":"https://openrouter.ai/api/v1/models"},"price_line":"$1.25 in, $4.25 out per 1M tokens","access":[{"provider":"Hugging Face (gated, manual approval)","model_id":"meta-models/Muse-Spark-Open-816B-A42B","url":"https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B"},{"provider":"OpenRouter","model_id":"meta/muse-spark-open","url":"https://openrouter.ai/meta/muse-spark-open"},{"provider":"Baseten, Fireworks AI, Groq, Modal, Nebius, Together AI (hosted APIs named by Meta)","docs":"https://research.meta.ai/blog/muse-spark-open"},{"provider":"Amazon SageMaker AI, Google Cloud Model Garden, Microsoft Foundry (weights)","docs":"https://research.meta.ai/blog/muse-spark-open"},{"provider":"Apple MLX (local, 2–4 M5 Ultra Mac Studios)","url":"https://github.com/ml-explore/mlx-lm","docs":"https://machinelearning.apple.com/research/mlx-distributed-inference"}],"capabilities":[{"name":"Largest US open-weight model at release","detail":"816B total / 42B active natively multimodal MoE with 1M context; Meta calls it 'the most capable US open-weight model to date' (own evaluation).","first":false,"discovered":"launch","source":"https://research.meta.ai/blog/muse-spark-open"},{"name":"Frontier-scale model on desktop machines","detail":"Runs in 4-bit NVFP4 (~493 GB) across two to four M5 Ultra Mac Studios linked by Thunderbolt 5 RDMA; about 25 tok/s on two and 37 tok/s on four (Apple ML Research).","first":true,"discovered":"launch","source":"https://machinelearning.apple.com/research/mlx-distributed-inference"}],"entry":"2026-10-15-meta-muse-spark-open","notes":"First Muse model with open weights. Knowledge cutoff inferred from the chat template default cutoff_date 2026-01-04 on Hugging Face (not stated in prose). Reasoning effort low/medium/high/xhigh (default xhigh). Benchmarks not yet independently reproduced.","verified":"2026-10-15","body_md":"Open-weight sibling of the hosted Muse Spark models; engines: SGLang, vLLM, TensorRT-LLM, Transformers, MLX, llama.cpp, Unsloth.\n</content>\n</invoke>","page_url":"https://postcutoff.com/m/muse-spark-open/","events_after":1194,"major_after":257,"historic_after":42,"missing_at_launch":1175,"events_since_release":null,"you_url":"https://postcutoff.com/you/muse-spark-open/","briefings":{"xs":"https://postcutoff.com/briefings/cutoff-2026-01-xs.md","s":"https://postcutoff.com/briefings/cutoff-2026-01-s.md","m":"https://postcutoff.com/briefings/cutoff-2026-01-m.md","full":"https://postcutoff.com/briefings/cutoff-2026-01.md"}}