Tool use
WinnerMCP Atlas
Scaled multi-tool / MCP use. Muse leads the listed set.
SPIRITT
Meta Muse 1.1Meta Muse 1.1 is built for agents, not chat
Meta Superintelligence Labs built Muse Spark 1.1 as a multimodal reasoning model for agentic tasks: tool use, computer use, coding, and multi-agent orchestration. On SPIRITT, you run it inside a fully equipped cloud environment with tools, browser, and durable memory.
Launch week on X
Real posts from people testing Muse, not launch-thread PR. Cost, agents, multimodal, and independent benches.
Official launch card
From Meta: Muse Spark 1.1 is a multimodal reasoning model built for agentic tasks, with major gains in tools, computer use, coding, and multimodal understanding. Now available to developers through the Meta Model API.
What Meta built it for
Planning and orchestration across apps and services. Act as a main agent, delegate to parallel subagents, generalize zero-shot to new tools and MCP servers, and keep long workflows coherent with a 1M token context plus active compaction.
Coding and computer use are first-class: diagnose bugs, ship features, migrate code, write scripts when automation is faster, click when UI interaction is simpler, and use vision to inspect what it just built.
Agentic orchestrationTool + MCPComputer useMultimodal1M contextMeta Model API
AA Intelligence Index
Source: Artificial Analysis (Jul 10, 2026). Overall ranking only. The charts below show the jobs Muse is built for.
Where Muse is strong vs other models
Same score means nothing alone. These are head-to-head charts so you can see where it leads, where it is close, and what that means for real agent work.
Professional work
WinnerJobBench
Professional tool-use tasks. Muse is the listed leader.
Finance agents
WinnerFinance Agent v2
Agentic financial analysis. Muse leads Opus and GPT.
Tool reasoning
WinnerHumanity's Last Exam (with tools)
Hard multidisciplinary reasoning with tools. Muse leads.
Computer use
Near frontierOSWorld-Verified
Real desktop computer use. Muse is near the top, slightly behind Opus.
Domain agents
WinnerHealthBench Professional
Professional health workflows. Muse leads the listed set.
Scientific coding
SciCode (AA)
Independent AA scientific coding. Muse ranks #3 overall.
Efficiency
Near frontierCost per AA Intelligence task
Lower is better. Muse is 2nd cheapest among the listed near-frontier options.
Sources: Meta Muse Spark 1.1 evaluation materials and Artificial Analysis public writeups. Competitive, not dominant, on pure coding boards like SWE-Bench Pro. The practical pitch is finishing agent workflows. On SPIRITT, you do not just call an API. You run Muse inside a workspace with tools, memory, and a computer already set up.
How It Works
From zero to Meta Muse 1.1 running real work in a cloud agent environment
01
Open a workspace
Start free, including trial. You land in a fully equipped cloud computer: browser, files, terminal, integrations, and memory. No local setup. No thin chat box pretending to be an agent.
02
Pick Meta Muse 1.1
Open the model picker and choose Meta Muse 1.1. Same model shipped for agentic work, now inside a workspace that already has tools, browser control, and durable context. Available across plans, including trial.
03
Build or automate
Tell it what to ship or what to run. Muse can code, call tools, drive the browser, coordinate multi-step work, and keep going while you step away. The point is not another chat window. It is an agentic environment where Muse actually does the job.
Try Muse 1.1 in an agentic environment
A real cloud environment, model picker, tools, and memory. Included in all plans, including trial.
Questions
Frequently asked questions
01What is Meta Muse 1.1?+
Meta Muse 1.1 (Muse Spark 1.1) is Meta Superintelligence Labs' multimodal reasoning model built for agentic tasks: tool use, computer use, coding, and multi-agent orchestration. It is available in Meta AI Thinking mode and through the Meta Model API, and you can run it inside SPIRITT Workspaces.
02How is Muse 1.1 different from previous generations?+
Muse Spark 1.1 is a step up from Muse Spark 1.0 on agentic work: stronger tool use, computer use, coding, and multi-agent orchestration, plus a 1M-token context with active compaction. It also opens through the public Meta Model API, so developers can run it outside Meta AI chat surfaces.
03What is Muse 1.1 best for?+
Agent workflows that plan, call tools, operate apps, and finish multi-step work. It stands out on MCP-style tool use, professional job tasks, finance agents, and tool-augmented reasoning. It is competitive on coding, not always the pure-coding leader.
04What are some of the main limitations of this model?+
It does not win every pure coding board (for example SWE-Bench Pro still favors Opus-class leaders). Long-context retrieval is mixed even with a 1M window. Multimodal scores are strong but not always best-in-class. Treat vendor and independent benches as signals, then test on your real workflow.
05What is a good example of using Muse 1.1 well?+
A real agent loop: open a workspace, pick Muse 1.1, hand it a multi-step goal (debug a UI from screenshots, migrate a module, research then ship a draft), and let it use tools, browser, and files until the job is done. That is where it beats single-turn chat demos.
06Where is Meta Muse 1.1 best to test for full agentic capabilities?+
SPIRITT Workspaces. Pick Meta Muse 1.1 in the model picker and run real work in a fully equipped cloud environment: tools, browser, files, terminal, memory, and durable sessions. Available on every plan, including trial. That is where the model can act as an agent, not just chat.
Buy from builders who use what they sellBuilt using