Meta jumps after it releases Superintelligence Labs’ first model: Muse Spark
The first big release from Meta’s Superintelligence Labs is here — a new multimodal reasoning model called Muse Spark. Shares of Meta spiked on the news, extending gains it had made earlier in the day on optimism over the ceasefire with Iran. The stock was recently up about 9%.
Meta has been playing catch-up in the generative-AI race, watching startups OpenAI and Anthropic leap ahead with ever more capable models, after the bungled rollout of its Llama 4 models.
After Meta went on an expensive hiring spree assembling an all-star team of AI researchers, investors have been eager to see the fruits of this team, and to see if the accompanying billions of capex dedicated to power it — $115 billion to $135 billion this year alone — were worth it.
Meta says the release is the first in a Muse family of models, which it says it will scale up over time. The benchmark scores released by Meta show Spark to be capable, with solid scores among popular benchmarks, but not any huge leaps over leading models from Anthropic, OpenAI, xAI, and Google.
Meta CEO Mark Zuckerberg said in a post on Threads:
“Looking ahead, we plan to release increasingly advanced models that push the frontier of intelligence and capabilities, including new open source models. We are building products that don’t just answer your questions but act as agents that do things for you. I am optimistic that this will support a wave of creativity, entrepreneurship, growth, and health. I’m looking forward to sharing more soon.”
After Meta went on an expensive hiring spree assembling an all-star team of AI researchers, investors have been eager to see the fruits of this team, and to see if the accompanying billions of capex dedicated to power it — $115 billion to $135 billion this year alone — were worth it.
Meta says the release is the first in a Muse family of models, which it says it will scale up over time. The benchmark scores released by Meta show Spark to be capable, with solid scores among popular benchmarks, but not any huge leaps over leading models from Anthropic, OpenAI, xAI, and Google.
Meta CEO Mark Zuckerberg said in a post on Threads:
“Looking ahead, we plan to release increasingly advanced models that push the frontier of intelligence and capabilities, including new open source models. We are building products that don’t just answer your questions but act as agents that do things for you. I am optimistic that this will support a wave of creativity, entrepreneurship, growth, and health. I’m looking forward to sharing more soon.”
