Long-running coding agents
Engineering workflows that require repository context, tool use, implementation steps and checks over a sustained session.
Independent tool overview
Muse Spark 1.3 at a glance
Muse Spark 1.3 is Meta’s model for long-horizon coding and agentic work, available through Muse Code and Meta Model API with standard and data-contributor pricing.
Visit the official Muse Spark 1.3 site ↗
- Best for
- Long-horizon coding and multi-step agents
- API model
- muse-spark-1.3
- Standard API price
- $1.25 input / $4.25 output per 1M tokens
- Access
- Muse Code and Meta Model API
- Open weights
- Announced for a future release, not available yet
Overview
What Muse Spark 1.3 is
Muse Spark 1.3 is designed to sustain longer tasks, coordinate multiple workflows in one thread and use tools to gather context, repair plan gaps and produce a final deliverable. Meta says it also follows long-form instructions more reliably than earlier Muse Spark releases.
For coding work, the model targets cleaner, less verbose execution and fewer unnecessary turns. It is available in Meta’s Muse Code terminal agent and through Meta Model API, where developers can choose standard data terms or a deeply discounted contributor tier.
Use cases
Who Muse Spark 1.3 is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Multi-workflow assistants
Agents that must keep several tasks straight inside one thread and map new or interrupting prompts back to the correct objective.
Structured knowledge work
Open-ended assignments that combine messy sources, planning, document creation and a final deliverable with many retained constraints.
Cost-sensitive prototyping
Teams willing to use Meta’s contributor data terms in exchange for much lower input and output token rates during experimentation.
Capabilities
Core Muse Spark 1.3 features
Long-horizon agent work
Uses tools to build context, revisit gaps in a plan and continue toward a final result across a larger open-ended task.
Multi-task thread handling
Targets messy conversations with multiple workflows, steering messages and interruptions without losing the active task.
Collaborative behavior
Trained to ask clarifying questions, request help when blocked and confirm before consequential actions.
Improved instruction retention
Designed to preserve detailed, long-form requirements across multi-step work with less constraint drift.
More efficient coding style
Meta reports fewer unnecessary turns, roughly 20% fewer tool calls and about 25% fewer tokens than Muse Spark 1.2 in comparisons by its engineers.
Muse Code and API deployment
Works through Meta’s terminal coding agent or as a model inside custom applications built on Meta Model API.
Process
How the Muse Spark 1.3 workflow works
Step 1
Choose Muse Code or the API
Install Muse Code for an interactive terminal agent, or select muse-spark-1.3 in Meta Model API for a custom product workflow.
Step 2
Set the data and pricing tier
Use Standard when prompts and completions must not train Meta’s models, or Contributor only when its training permission is acceptable.
Step 3
Define the objective and permissions
Provide detailed requirements, available tools and explicit boundaries for any file, account or external action the agent may take.
Step 4
Monitor milestones and verify output
Review the model’s plan, test generated code and independently confirm claims or consequential changes before acceptance.
Cost
Muse Spark 1.3 pricing and free plan
Meta Model API offers Muse Spark 1.3 in Standard and Contributor tiers. Contributor pricing is dramatically lower because it permits Meta to use prompts and completions to train future models; teams should treat that data choice as part of the price decision.
Standard API tier
$1.25 input · $4.25 output / 1M tokens
Standard usage pricing with prompts and completions excluded from Meta model training.
- $0.15 per million cached-input tokens
- No long-context price premium
- Web-search grounding is charged separately when used
Contributor API tier
$0.10 input · $0.20 output / 1M tokens
Discounted pricing in exchange for permission to use prompts and completions to train future Meta models.
- $0.002 per million cached-input tokens
- Uses model identifier muse-spark-1.3-contributor
- Appropriate only when the training-data terms fit the workload
Pricing checked . Check current pricing at the source ↗
Assessment
Muse Spark 1.3 strengths and limitations
Where it stands out
- Designed around sustained agent workflows, instruction retention and tool use rather than only short responses.
- Available both as a coding agent in the terminal and as an API model for custom applications.
- Standard token prices are competitive for a model positioned around coding and long-horizon work.
- Contributor pricing offers unusually low prototype costs when its data-training terms are acceptable.
What to consider
- Max reasoning was not available at launch; Meta said it would follow after additional safety testing.
- The announced open-weights release is still on Meta’s roadmap rather than part of Muse Spark 1.3’s current availability.
- Contributor-tier prompts and completions may be used to train future Meta models, which can make that tier unsuitable for sensitive data.
- Meta’s efficiency figures come from its own engineering comparisons and should be validated on the team’s actual repositories and acceptance tests.
Compare
Muse Spark 1.3 alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Consumer
Claude Fable 5.1
Choose Claude Fable 5.1 when Claude Code, Cowork or Anthropic’s supported cloud deployments are more important than Meta’s pricing options.
Explore Claude Fable 5.1 →Consumer
Gemini 3.8 Flash
Choose Gemini 3.8 Flash when Google’s AI Studio, Enterprise and Workspace ecosystem or its launch input price is the better fit.
Explore Gemini 3.8 Flash →Consumer
GPT 5.5
Choose GPT 5.5 when existing OpenAI integrations and operational tooling outweigh the benefits of moving to Meta Model API.
Explore GPT 5.5 →Questions
Muse Spark 1.3 FAQs
What is Muse Spark 1.3?
Muse Spark 1.3 is Meta’s model for long-horizon coding, agentic workflows and multi-step knowledge work. It is available in Muse Code and Meta Model API.
How much does Muse Spark 1.3 cost?
Standard API pricing is $1.25 per million input tokens and $4.25 per million output tokens. Cached input is $0.15 per million tokens.
What is the Muse Spark contributor tier?
It is a discounted API tier priced at $0.10 per million input tokens and $0.20 per million output tokens. In exchange, Meta can use prompts and completions to train future models.
Does Muse Spark 1.3 work in Muse Code?
Yes. Meta launched Muse Spark 1.3 in Muse Code for terminal-based coding work and in Meta Model API for custom integrations.
Is Muse Spark 1.3 open source or open weights?
Not at the time of this review. Meta described an open-weights Muse Spark release as part of its future roadmap.
Does Muse Spark 1.3 support max reasoning?
Meta said previously available reasoning modes were ready at launch, with max reasoning planned shortly afterward following additional safety testing.
Is this a hands-on Muse Spark 1.3 review?
No. This independent overview uses Meta’s launch, evaluation and pricing documentation; The Rundown has not completed a controlled hands-on model comparison for this page.
Bottom line
Our Muse Spark 1.3 verdict
Muse Spark 1.3 is a credible option for coding and agent teams that want long-horizon behavior at competitive API rates, particularly if Muse Code fits their terminal workflow. Its contributor tier is inexpensive but creates a consequential data-use tradeoff, and launch claims still need workload-specific testing. Claude Fable 5.1 and Gemini 3.8 Flash are the clearest comparisons for teams deciding among ecosystem, capability and cost.
Visit Muse Spark 1.3 website ↗Similar Consumer Tools
View all Consumer AI tools →
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.