441e85161ClaimMeta releases Muse Spark Open, an 816-billion-parameter model it calls the most capable US open-weight AI modelAgreeDisagree
441e85161ClaimAt its October 2026 launch, Meta's Muse Spark Open was free to use under the Muse Model License, which requires products to be labeled "Built with Muse Spark Open"AgreeDisagree
441e85161ClaimAt its October 2026 launch, Meta called Muse Spark Open the most capable US open-weight AI model to dateAgreeDisagree
441e85161ClaimMeta's 816-billion-parameter Muse Spark Open can run locally on two Apple M5 Ultra Mac Studios using Apple's MLX frameworkAgreeDisagree
441e85161ClaimMeta's Muse Spark Open has 816 billion total parameters, 42 billion of them active, and a 1 million token context windowAgreeDisagree
441e85161ClaimMeta said on October 12, 2026 that it would release open weights for Muse Spark but not for its larger Muse Wonder models, citing their size and potential safety concernsAgreeDisagree
441e85161ClaimMeta said in October 2026 that Muse Spark Open was the first model of its class able to run on a desktop setupAgreeDisagree
441e85161ClaimMeta released Muse Spark Open's weights on Hugging Face, Amazon SageMaker AI, Google Cloud Model Garden and Microsoft Foundry ModelsAgreeDisagree
441e85161ClaimUnder Meta's October 2026 Muse Model License, API providers earning more than $20 million a year need a separate agreement with Meta to offer Muse Spark Open as a serviceAgreeDisagree
441e85161ClaimAt launch in October 2026, Meta's open-weight Muse Spark Open fell well short of Meta's paid Muse Wonder Full model on benchmarks, according to The DecoderAgreeDisagree
441e85161ClaimMeta gave Muse Spark Open additional safety post-training because the model's weights are publicAgreeDisagree
441e85161ClaimMeta said in October 2026 that Muse Spark Open met the release thresholds of Meta's Superintelligence Scaling FrameworkAgreeDisagree
441e85161ClaimMeta's Muse Spark Open scored 39 on the Artificial Analysis Intelligence Index in October 2026, well above the median of 19 for comparable modelsAgreeDisagree
441e85161ClaimMeta reported in October 2026 that Muse Spark Open beats Alibaba's Qwen 3.8, Z.ai's GLM 5.3 and Moonshot AI's Kimi K3 in agentic, knowledge and multimodal benchmarks but trails some rivals in codingAgreeDisagree
441e85161ClaimChina is advancing the open-weight AI model industry in a way the US is notAgreeDisagree