Alexandr Wang on X: "1/ big announcement today: we will be releasing an open weight version of muse spark 1.2 soon.
we also are releasing muse glimmer, a 30B agentic model with open weights under apache 2.0. muse glimmer can run on 24GB of VRAM without losing agentic reliability. 馃У" / X<br>Post
Log inSign up
Post
Alexandr Wang
@alexandr_wang
1/ big announcement today: we will be releasing an open weight version of muse spark 1.2 soon.
we also are releasing muse glimmer, a 30B agentic model with open weights under apache 2.0. muse glimmer can run on 24GB of VRAM without losing agentic reliability. 馃У<br>span:not(:empty)~span:not(:empty)]:before:content-['路'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0">10:06 AM 路 Aug 10, 202618.2KViews
52<br>62<br>589<br>67
span:not(:empty)~span:not(:empty)]:before:content-['路'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">Alexandr Wang
@alexandr_wang
28m
2/ just like much larger models, muse glimmer can operate as a fully capable agent via planning, tool calls, checking its own results, and failure recovery.
00:00
74<br>4.1K
span:not(:empty)~span:not(:empty)]:before:content-['路'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">Alexandr Wang
@alexandr_wang
28m
3/ muse glimmer was developed with its own architecture and recipe, optimized for its size and agentic performance requirements.
59<br>3.3K
span:not(:empty)~span:not(:empty)]:before:content-['路'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">Alexandr Wang
@alexandr_wang
28m
4/ to achieve maximum memory efficiency, we quantize model weights to ~4-bit, getting the language model under 20GB with room for the kv cache, perception encoder, and drafter alongside it. a dflash drafter proposes blocks of tokens the main model verifies in parallel, so it Show more
00:00
33<br>2K
span:not(:empty)~span:not(:empty)]:before:content-['路'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">Alexandr Wang
@alexandr_wang
28m
5/ weights on hugging face now. running this week through ollama, LM Studio, vllm, sglang, together, fireworks, and openrouter, with llama.cpp, MLX, and executorch.
evaluated under meta's advanced AI scaling framework and cleared for open-weight release.
42<br>3.3K
span:not(:empty)~span:not(:empty)]:before:content-['路'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">Alexandr Wang
@alexandr_wang
28m
6/ that's the short version. the full write-up, including evals and the training details, is here:
Introducing Muse Glimmer: An Open Agentic Model That Runs on Your Device
From research.meta.ai
47<br>3.1K
span:not(:empty)~span:not(:empty)]:before:content-['路'] [&>span:not(:empty)~span:not(:empty)]:before:px-1 [&>span:not(:empty)~span:not(:empty)]:before:shrink-0 min-w-0 overflow-hidden">sensho
@sensho
13m
goated release and ty for making meta release os weights again but SURELY there's a better time to announce this than 3am
539
Log in or sign up for X<br>See what鈥檚 happening and join the conversation<br>Continue with phoneContinue with AppleContinue with Google<br>or<br>Log in with username or email
Relevant people
Alexandr Wang@alexandr_wangFollow<br>chief ai officer @meta, founder meta superintelligence labs, founder @scale_ai. rational in the fullness of time
Trending now