Muse Spark
Reasoning modelMuse Spark by Meta Superintelligence Labs
Meta Muse Spark 是 Meta Superintelligence Labs 于 2026 年 4 月发布的首款模型，也是 Llama 4 失利后 Meta 全面重建 AI 研发体系的第一个对外成果。模型由首席 AI 官 Alexandr Wang 领导团队历时九个月开发完成，原生支持多模态输入，内置多智能体并行推理机制。基准测试中，Muse Spark 在医疗问答（HealthBench Hard 42.8%）和图表理解（CharXiv Reasoning 86.4）上表现突出，整体推理和智能体编码能力与 GPT-5.4、Gemini 3.1 Pro 仍有差距。Meta 将其定位为 Muse 系列的起点，更大规模的后续模型已在开发中。
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
Model basics
Open source & experience
Official resources
API details
Benchmark Results
Muse Spark currently shows benchmark results led by HLE (4 / 159, score 58), GPQA Diamond (23 / 179, score 89.50), FrontierMath (9 / 60, score 39). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.
General Knowledge
5 evaluationsCoding and Software Engineer
1 evaluationsMath and Reasoning
3 evaluationsCompare with other models
Publisher
Model Overview
Meta Muse Spark 是 Meta Superintelligence Labs 于 2026 年 4 月发布的首款模型，也是 Llama 4 失利后 Meta 全面重建 AI 研发体系的第一个对外成果。模型由首席 AI 官 Alexandr Wang 领导团队历时九个月开发完成，原生支持多模态输入，内置多智能体并行推理机制。基准测试中，Muse Spark 在医疗问答（HealthBench Hard 42.8%）和图表理解（CharXiv Reasoning 86.4）上表现突出，整体推理和智能体编码能力与 GPT-5.4、Gemini 3.1 Pro 仍有差距。Meta 将其定位为 Muse 系列的起点，更大规模的后续模型已在开发中。
DataLearner on WeChat
Follow DataLearner on WeChat for AI model updates and research notes.
