Agent developers
Prototype applications that plan, call functions, maintain long context, and work across external tools.
Independent tool overview
Muse Spark 1.1 at a glance
Muse Spark 1.1 is Meta's July 2026 multimodal reasoning model for agentic workflows, computer use, coding, and long-context tasks; Muse Spark 1.2 is now the newer release.
Visit the official Muse Spark 1.1 site ↗
- Developer access
- Meta Model API public preview
- Consumer access
- Meta AI Thinking mode
- Context window
- Up to 1 million tokens
- Newer version
- Muse Spark 1.2
- Last reviewed
- August 30, 2026
Overview
What Muse Spark 1.1 is
Muse Spark 1.1 is a hosted multimodal reasoning model from Meta Superintelligence Labs. Released in July 2026, it expanded the original Muse Spark with stronger tool use, computer interaction, coding, multimodal understanding, and active management of a 1 million-token context window.
Meta made the model available in Thinking mode through Meta AI and opened a public preview of the Meta Model API for developers. Its agentic design supports long-running workflows that plan, call tools, work across interfaces, and coordinate parallel subagents, while its multimodal inputs include images, video, audio, and documents.
Muse Spark 1.1 is no longer Meta's latest Spark release. Meta introduced the coding-optimized Muse Spark 1.2 and Muse Code in August 2026, so new production evaluations should compare the current model catalog instead of defaulting to 1.1 solely because of its original launch claims.
Use cases
Who Muse Spark 1.1 is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Computer-use workflows
Test agents that combine scripted automation with direct interaction across changing user interfaces.
Large-codebase tasks
Evaluate bug fixes, feature work, migrations, visual debugging, and multi-step coding workflows.
Multimodal projects
Reason over combinations of text, screenshots, images, video, audio, PDFs, and tool results.
Capabilities
Core Muse Spark 1.1 features
Long active context
The model manages a context window of up to 1 million tokens, retrieving earlier work and compacting history during extended tasks.
Tool and function calling
The Meta Model API exposes agentic affordances for connecting model reasoning to developer-controlled tools and services.
Computer use
Muse Spark 1.1 can switch between writing automation scripts and operating interfaces directly, including batching related actions.
Multi-agent orchestration
It was trained to act as a coordinating main agent or a bounded subagent and to escalate when a delegated job requires help.
Agentic coding
Meta positions the model for debugging, feature implementation, migrations, visual inspection, and multi-turn work in large codebases.
Multimodal perception
It can incorporate visual, audio, video, and document information into workflows that also take actions or call tools.
Process
How the Muse Spark 1.1 workflow works
Step 1
Check the current model
Compare Muse Spark 1.1 with Muse Spark 1.2 and other current options before pinning an older model ID in a new application.
Step 2
Build a representative evaluation
Use real task shapes, allowed data, tool schemas, expected outputs, and failure cases rather than relying only on vendor benchmarks.
Step 3
Constrain the tools
Give the model the minimum permissions needed, use strict allowlists, isolate workspaces, and require approval for consequential actions.
Step 4
Measure the full agent loop
Track success rate, retries, tool errors, latency, long-context drift, token usage, human interventions, and total cost per completed task.
Step 5
Monitor production behavior
Log decisions and actions, defend against prompt injection, set spending and action limits, and maintain a reliable human override path.
Cost
Muse Spark 1.1 pricing and free plan
Muse Spark 1.1 is a hosted model rather than an open-weight download. Meta Model API pricing is usage based; the reviewed public rate is $1.25 per million input tokens, $0.15 per million cached input tokens, and $4.25 per million output tokens. Search grounding and third-party gateways can add separate charges, and preview pricing can change.
Meta AI
Consumer access
Muse Spark 1.1 launched in Thinking mode in the Meta AI app and on meta.ai; consumer availability and limits follow Meta's current product access.
- No model API key required
- Not a self-hosted deployment
- Features and limits can vary by account and region
Meta Model API
$1.25 input / $4.25 output per 1M tokens
Pay-as-you-go developer access during the Meta Model API public preview.
- $0.15 per 1M cached input tokens
- Hosted API; model weights are not included
- Tool, search, gateway, and application costs may be separate
Pricing checked . Check current pricing at the source ↗
Assessment
Muse Spark 1.1 strengths and limitations
Where it stands out
- Combines long context, multimodal perception, tool use, computer interaction, and coding in one hosted model.
- Designed for agent loops that coordinate multiple steps and delegated subagents rather than only single-turn answers.
- Usage pricing is competitive for teams that can keep retries and tool-loop overhead under control.
- Meta publishes a detailed evaluation report covering capabilities, adversarial robustness, model behavior, and deployment risk.
What to consider
- Muse Spark 1.2 is now the newer release, making 1.1 a version to maintain or compare rather than the automatic choice for new builds.
- The hosted API does not provide open weights or a self-hosting path for Muse Spark 1.1.
- A 1 million-token window can increase cost and still does not guarantee perfect recall, prioritization, or consistency across a long task.
- Computer-use and tool-calling agents remain vulnerable to bad instructions, indirect prompt injection, excess permissions, and irreversible actions without application-level controls.
- Public-preview availability, model behavior, pricing, limits, and regional access can change.
Compare
Muse Spark 1.1 alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Consumer
Gemini 3.1 Pro
Compare Google's flagship multimodal reasoning model for long-context, tool-use, and enterprise application workloads.
Explore Gemini 3.1 Pro →Consumer
Claude Opus 4.8
A strong alternative for demanding coding and agentic workflows where reliability and extended task execution are priorities.
Explore Claude Opus 4.8 →Consumer
Grok 4.5
Consider xAI's current model when evaluating agentic performance, coding capability, speed, and API cost across providers.
Explore Grok 4.5 →Questions
Muse Spark 1.1 FAQs
What is Muse Spark 1.1?
Muse Spark 1.1 is Meta's July 2026 hosted multimodal reasoning model for agentic tasks, tool use, computer interaction, coding, and long-context workflows.
Is Muse Spark 1.1 the latest version?
No. Meta introduced Muse Spark 1.2 and Muse Code in August 2026. Check the current model catalog before starting a new integration.
How large is its context window?
Meta describes a 1 million-token context window with active context management that retrieves earlier work and compacts history during long tasks.
Can I self-host Muse Spark 1.1?
Meta's 1.1 release is provided through Meta AI and a hosted Model API. The release does not include open model weights for self-hosting.
How much does the API cost?
The reviewed direct rate is $1.25 per million input tokens, $0.15 per million cached input tokens, and $4.25 per million output tokens. Confirm current preview pricing and any separate tool or search charges before deployment.
Is computer use safe without supervision?
No agent should receive broad unsupervised access by default. Use minimal permissions, tool allowlists, isolated workspaces, action and spending limits, logs, and human approval for consequential operations.
Bottom line
Our Muse Spark 1.1 verdict
Muse Spark 1.1 was an important step in Meta's move from a consumer assistant model to a developer-facing agentic platform. Its multimodal, long-context, coding, and computer-use combination remains technically relevant, but new projects should evaluate Muse Spark 1.2 and current competitors while measuring complete task success, safety controls, and cost—not model pricing in isolation.
Visit Muse Spark 1.1 website ↗Similar Consumer Tools
View all Consumer AI tools →
GPT-Live
GPT-Live - OpenAI's new voice model with more natural conversation abilities
ChatGPT Work
ChatGPT Work- OpenAI's Codex-powered agent for everyday non-coding work
Willow Frontier Mini
Willow Frontier Mini - Willow's free, unlimited AI dictation model
Bonsai 27B
Bonsai 27B - PrismML's compressed 27B model small enough for iPhones
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.