Meta Shares Muse Spark 1.2 Multimodal Benchmarks
Meta AI Research posted benchmarks and demos for the model's visual and robotics tasks.
TLDR
Meta AI Research published a blog post on the multimodal intelligence of Muse Spark 1.2. Company posts describe performance across visual reasoning and chart understanding benchmarks plus code generation from image or video inputs. A variant serves as a robot brain that decodes instructions, calls tools, observes outcomes, and iterates until tasks finish. Demos illustrate bimanual robot planning and agentic media generation. Alexandr Wang and AI at Meta accounts posted the updates along with photos and videos.
Combined views
358.6K
15 Sources, first seen ago
Sources
XW 
Xiaolong Wang@xiaolonwMuse Spark, now making its way onto robots. Some early results of our work at Meta 👀 https://twitter.com/alexandr_wang/status/2090485261629354352
- likes: 122
- replies: 5
- bookmarks: 21
- reposts: 13
LP Lerrel Pinto@LerrelPintoStrong multimodal models = strong robot models. Check out some early glimpses of Meta's Muse Spark 1.2 on robots here! https://twitter.com/alexandr_wang/status/2090485261629354352
- likes: 74
- replies: 4
- bookmarks: 11
- reposts: 9
AA AI at Meta@AIatMetaToday we’re also previewing WildArtifactBench, an internal evaluation framework designed to assess agents on complex, real-world tasks across diverse deliverable formats. By using win rates and Elo scores from human and agentic preference judges rather than strict ground-truth…
- likes: 569
- replies: 68
- bookmarks: 155
- reposts: 55
WC Wenhu Chen@WenhuChenTry our Muse 1.2 for multimodal tasks!Big shout out to the MSL MM team. https://twitter.com/alexandr_wang/status/2090485261629354352
- likes: 35
- replies: 2
- bookmarks: 7
- reposts: 2
CA Capitolino@soookidaPlease help me get my Facebook account back. I genuinely don’t believe I did anything wrong, but my account was suddenly disabled. The biggest problem is that Meta seems to have no real human customer support available. Before investing so heavily in AI, maybe Meta should have…
VY Volkan Yürülmez@VolkanYrlmez@WenhuChen I think the breadth here is really impressive, visual coding and robotics and audio-visual understanding all working together. Found a solid explainer on it: https://x.com/AGTPinsights/status/2090497781878214873