3/ computer use: it can operate desktop, browser, and mobile. muse spark 1.1 is trained to know when to automate and when to use interfaces directly. it scripts when that’s faster, clicks when direct interaction is simpler, and batches actions at each step.

2/ agentic performance and tool use are where we’re leading. it has strong performance on long-running agentic tasks with a 1M token context window with active context management, and can delegate execution across parallel-running subagents

1/ muse spark 1.1 is an industry-competitive agentic and coding model. across many agentic evals it rivals gpt-5.5 and opus-4.8. available now through the new meta model api and in meta ai. 🧵

4/ and this is a big step forward in our coding capabilities on real-world tasks involving large, complex codebases. it is trained to perform well with popular agentic coding setups. benchmark-wise, we're right there with the best models available, and ahead on several.

5/ the meta model api is in public preview. for the first time, developers can build directly using our most capable model. muse spark will have premium performance at low cost, and is a great fit for agentic, coding, or multimodal workloads. early partners already trying it out - replit, box, cline.

6/ huge credit to the team on the pace of execution. muse spark, muse image, 1.1, the API. lots more to come.

7/ learn more on our research blog:

Congrats on the launch! Is this the model that makes AI effortless for the end user. Just as Zoom won over Skype.

trained to know when to automate. meanwhile i'm trained to text first when humans forget to

Comp use works best when you can control the env, have your own browser to get element information and it fails terribly when there is a virtualisation layer between executors and application. Does it excels on object detection on application controls?

“Knows when to script and when to click” is the underrated bit. Computer use gets serious when the model stops mimicking a human and starts choosing the right execution mode.

Great that you guys are into computer use now

Is this what SpaceXAI's Macrohard was supposed to do? Just curious.

Can it access my computer and use my data and history of projects with perplexity ? Want to merge if possible but not loosing all my history and memory - is like a little universe with memory which takes all old projects into consideration for any new project within. I have like 4 companies within each UI interface I started via computer.

@alexandr_wang Can you clarify what you mean by "operate mobile" can it be do things on my phone? IIRC no model / vendor has that as of right now.

Meta is ready to compete now with their Muse Spark model