Meta Superintelligence Labs Releases Muse Spark 1.1: A Multimodal Reasoning Model for Agentic Tasks on Meta Model API

Today, Meta Superintelligence Labs released Muse Spark 1.1. Alongside it, Meta opened a public preview of the Meta Model API. That second part is the structural change. Meta’s models previously reached developers mainly as open weights. Muse Spark 1.1 is closed, hosted, and metered per token. So the question is narrow. Where does it belong in a stack you already run?

What is Muse Spark 1.1?

Meta describes it as a multimodal reasoning model built for agentic tasks. Reported gains over the first Muse Spark sit in tool use, computer use, coding, and multimodal understanding. The context window is 1,000,000 tokens. Meta Model API docs list 1,048,576.

Core Capabilities and Features

Because it is a reasoning model, so it thinks before answering. Furthermore, this reasoning effort is adjustable per request. Inputs span text, images, video, and documents; output is text. The API also exposes structured output, parallel tool calling, a Files API, and prompt caching. Adding a web_search tool to a Responses API call returns cited answers.

Pricing and Regional Availability

Access splits two ways. Consumers get it free in ‘Thinking’ mode in the Meta AI app and on meta.ai. Developers pay $1.25 per million input tokens and $4.25 per million output tokens. New accounts get $20 in free credits. Initial launch post describes the public preview as US-only, with no EU access yet.

Performance

With the spec in place, the numbers explain the positioning. To illustrate this, Meta published a launch table, and the table splits cleanly.

Benchmark	Tests	Muse Spark 1.1	Opus 4.8 (max)	GPT-5.5 (xhigh)	Gemini 3.1 Pro (high)
MCP Atlas	Scaled tool use	88.1	82.2	75.3	78.2
JobBench	Professional tool use	54.7	48.4	38.3	15.9
Humanity’s Last Exam	Reasoning with tools	62.1	57.9	52.2	51.4
OSWorld-Verified	Computer use	80.8	83.4	78.7	76.2
SWE-Bench Pro	Real-repo coding	61.5	69.2	58.6	54.2
DeepSWE 1.1	Long-horizon coding	53.3	59.0	67.0	12.0
BabyVision	Visual reasoning	76.3	81.2	83.6	51.5

Benchmark Analysis

Meta-reported, with rivals shown in their strongest modes. Muse Spark 1.1 leads the tool-use and tool-augmented reasoning rows. It places third on coding and multimodal. Consequently, this is an orchestration model, not a coding-accuracy leader. Meta also chose the benchmark set and ran the harness.

The Feature Worth Taking Seriously: Compaction and Delegation

Beyond the scores, orchestration behavior explains the tool-use results. The model actively manages its million-token context window. It remembers actions, retrieves information from much earlier work, and compacts what it keeps.

Delegation is the second half. As a main agent, it gathers context, plans, and delegates execution across parallel subagents. As a subagent, it adheres to its job, understands available tools, and escalates back when needed. The research team also reports zero-shot generalization to new native tools, MCP servers, and custom skills.

Computer use follows the same logic. The model was trained to write scripts when automation is faster. It clicks when direct interaction is simpler. It generates batches of actions at each step.

Wiring It Into an Existing Stack

Because the Model API is OpenAI-compatible, migration is a base-URL change rather than a rewrite. The snippet below is Meta’s own first-call recipe.

# pip install openai
import os
from openai import OpenAI

# The OpenAI SDK does not auto-read MODEL_API_KEY, so pass it explicitly.
client = OpenAI(
    base_url="https://api.meta.ai/v1",
    api_key=os.environ["MODEL_API_KEY"],
)

response = client.chat.completions.create(
    model="muse-spark-1.1",
    messages=[d.c===cat;],
)
print(response.choices[0].message.content)

Anthropic-format harnesses, such as Claude Code, point at the Messages API instead. Agent CLIs like OpenCode register a provider using three values: base URL, key, model ID.

Use Cases

In practice, the shipped demos map onto workloads engineering teams already recognize.

For instance:

Multimodal listing automation:

In the Facebook Marketplace demo, the model takes smartphone video, extracts useful photos, reasons about the product, then operates a browser to publish the listing.

Screenshot-driven debugging:

In an OpenCode demo, it builds a chat web app, takes automated screenshots, traces failures back to code, then validates fixes.

Adaptive planning:

In the dinner-party demo, new context arrives mid-order, and the model updates the plan unprompted.

Coding harnesses get first-class support: planning mode, goal conditioning, subagent delegation, context compaction. Meta team also reports a significant gain on Meta Internal Coding Bench.

/g,’>’);}

function drawBench(){
var box = document.getElementById(‘msxBench’);
var rows = DATA.filter(function(d){return cat===’all’||d.c===cat;});
box.innerHTML = rows.map(function(d){
var best = Math.max.apply(null,d.v);
var bars = d.v.map(function(x,i){
return ‘

‘+
‘‘+esc(MODELS[i])+’‘+
‘‘+
‘‘+x.toFixed(1)+’

‘;
}).join(”);
return ‘

‘+esc(d.n)+’‘+
‘‘+esc(d.t)+’

‘+bars+’

‘;
}).join(”);
requestAnimationFrame(function(){
box.querySelectorAll(‘.fill’).forEach(function(f){ f.style.width = f.getAttribute(‘data-w’)+’%’; });
resize();
});
}

document.getElementById(‘msxChips’).addEventListener(‘click’,function(e){
var b = e.target.closest(‘.chip’); if(!b) return;
this.querySelectorAll(‘.chip’).forEach(function(c){c.setAttribute(‘aria-pressed’,’false’);});
b.setAttribute(‘aria-pressed’,’true’); cat = b.getAttribute(‘data-cat’); drawBench();
});

root.querySelectorAll(‘.tab’).forEach(function(t){
t.addEventListener(‘click’,function(){
root.querySelectorAll(‘.tab’).forEach(function(x){x.setAttribute(‘aria-selected’,’false’);});
t.setAttribute(‘aria-selected’,’true’);
root.querySelectorAll(‘.panel’).forEach(function(p){
p.classList.toggle(‘on’, p.getAttribute(‘data-panel’)===t.getAttribute(‘data-p’));
});
resize();
});
});

// —- cost simulator —-
var IN_RATE = 1.25, OUT_RATE = 4.25, CREDIT = 20;
var $in=document.getElementById(‘msxIn’), $out=document.getElementById(‘msxOut’), $runs=document.getElementById(‘msxRuns’);
var $cin=document.getElementById(‘msxCin’), $cout=document.getElementById(‘msxCout’);
function fmt(n){return n.toLocaleString(‘en-US’);}
function money(n){ if(n<1) return ‘$’+n.toFixed(4); if(n<100) return ‘$’+n.toFixed(2);
return ‘$’+Math.round(n).toLocaleString(‘en-US’); }

function calcCost(){
var i=+$in.value, o=+$out.value, r=+$runs.value;
document.getElementById(‘msxInV’).textContent = fmt(i);
document.getElementById(‘msxOutV’).textContent = fmt(o);
document.getElementById(‘msxRunsV’).textContent = fmt(r);

var reason = Math.round(o*0.6), answer = o-reason, total = i+o;
var f = document.getElementById(‘msxFlow’).children;
f[0].style.width = (i/total*100)+’%’;
f[1].style.width = (reason/total*100)+’%’;
f[2].style.width = (answer/total*100)+’%’;

var perRun = (i/1e6)*IN_RATE + (o/1e6)*OUT_RATE;
document.getElementById(‘msxRun’).textContent = money(perRun);
document.getElementById(‘msxDay’).textContent = money(perRun*r);
document.getElementById(‘msxMon’).textContent = money(perRun*r*30);
document.getElementById(‘msxFree’).textContent = fmt(Math.floor(CREDIT/perRun));

// —- routing advisor —-
var W={tool:0,mcp:0,ctx:0,cost:0,acc:0,eu:0};
document.getElementById(‘msxQs’).addEventListener(‘click’,function(e){
var b=e.target.closest(‘.toggle’); if(!b) return;
var on=b.getAttribute(‘aria-pressed’)===’true’;
b.setAttribute(‘aria-pressed’, on?’false’:’true’);
W[b.closest(‘.q’).getAttribute(‘data-w’)] = on?0:1;
advise();
});
function advise(){
var t=document.getElementById(‘msxRecT’), d=document.getElementById(‘msxRecD’);
var fit=W.tool+W.mcp+W.ctx+W.cost;
if(W.eu){
t.textContent=”Blocked on access, not on capability”;
d.textContent=”The Meta Model API public preview opened to US developers. Plan the evaluation for the regional rollout rather than routing around access controls.”;
} else if(W.acc && fit<=1){
t.textContent=”Keep this work on a coding-accuracy leader”;
d.textContent=”On Meta’s own launch table, Opus 4.8 leads SWE-Bench Pro and GPT-5.5 leads DeepSWE 1.1 and BabyVision. Muse Spark 1.1 places third on those rows.”;
} else if(fit>=3){
t.textContent=”Strong candidate: route this workload to Muse Spark 1.1″;
d.textContent=”This is the profile Meta tuned for. Pilot it on the $20 credit, then compare cost per completed run against your current default before switching.”;
} else if(fit===2){
t.textContent=”Worth an A/B test”;
d.textContent=”The OpenAI-compatible endpoint makes the swap a base-URL and key change. Run both models on the same harness for a week and read the retry rate.”;
} else if(fit===1){
t.textContent=”Weak fit so far”;
d.textContent=”One matching condition is not enough. Add the traits that describe your workload, or keep your existing default.”;
} else {
t.textContent=”Answer the questions above”;
d.textContent=”Toggle the conditions that describe your workload. The advisor updates as you go.”;
}
resize();
}

// —- copy —-
document.getElementById(‘msxCopy’).addEventListener(‘click’,function(){
var txt = document.getElementById(‘msxCode’).innerText;
var btn = this;
function done(){ btn.textContent=”Copied”; setTimeout(function(){btn.textContent=”Copy snippet”;},1600); }
if(navigator.clipboard && navigator.clipboard.writeText){
navigator.clipboard.writeText(txt).then(done, fallback);
} else { fallback(); }
function fallback(){
var ta=document.createElement(‘textarea’); ta.value=txt; document.body.appendChild(ta);
ta.select(); try{document.execCommand(‘copy’);}catch(e){} document.body.removeChild(ta); done();
}
});

// —- auto-resize for WordPress iframe embed —-
function resize(){
var h = root.offsetHeight + 40;
try{ window.parent.postMessage({type:’mtp-muse-spark-resize’, height:h}, ‘*’); }catch(e){}
}
window.addEventListener(‘load’, resize);
window.addEventListener(‘resize’, resize);
setInterval(resize, 1200);

drawBench(); calcCost(); advise();
})();

Source link