Muse Spark 1.3 Contributor

Overview

Muse Spark 1.3 Contributor is Meta’s lower-cost contributor tier for a multimodal reasoning model. It is positioned for experimentation, learning, and early-stage work involving agents, multiple agents, and coding. The summary says it is designed to track information, though it does not specify what information or how that capability works.

The model accepts text, images, video, and files, and returns text. It supports reasoning and tools, and has a context window of 1,048,576 tokens. Its model ID is meta/muse-spark-1.3-contributor. Meta released it on September 2, 2026.

For builders choosing a model to fit an existing stack, its broad input modalities and large context window are relevant selection points. The facts do not describe tool interfaces, deployment options, or specific coding and agent capabilities, so those details need to be evaluated separately.

Key features

Multimodal input

Text, image, video, and file inputs are supported, while output is text. Vision is listed as available. This makes the model’s stated input range broader than text alone, but no image or video processing limits are specified.

Reasoning and tools

Reasoning and tools are both listed as supported. That aligns with the model’s stated focus on early-stage agentic and multi-agent workflows. The facts do not identify particular tools, calling formats, or orchestration features.

Context and output

The context window is 1,048,576 tokens, and the maximum output is 943,718 tokens. These are separate limits: the listed maximum output is not the same as the context window. No further information is given about how those limits apply to individual requests.

Pricing

Input costs USD 0.1 per 1M tokens. Output costs USD 0.2 per 1M tokens. Output tokens are priced at twice the input rate, so teams estimating usage should account for both sides of a request. No other fees or pricing terms are listed.

Platforms

The facts identify the model and its input and output modalities, but do not specify supported platforms, APIs, SDKs, hosting arrangements, or deployment requirements. Teams should confirm that the available integration path fits their stack before making implementation plans.

Who it's for

Muse Spark 1.3 Contributor is described for experimentation and learning, as well as early-stage agentic, multi-agent, and coding workflows. Its listed reasoning and tool support may make it worth considering for teams exploring those areas, while its multimodal inputs may suit projects that need to pass more than text to a model.

The contributor designation and stated focus on early-stage use are important boundaries: the facts do not establish production readiness or provide performance guarantees. Teams evaluating it for more demanding workloads will need details beyond the capabilities and limits listed here.

Pros and cons

• Pros: Supports text, image, video, and file inputs; includes listed reasoning and tool support; provides a context window of 1,048,576 tokens.
• Cons: The available facts do not explain tool interfaces, modality-specific limits, integration options, or performance. Output tokens cost more than input tokens.

Alternatives

For a comparison within Meta’s model catalog, consider Muse Spark 1.3, Muse Spark 1.2 Contributor, Muse Glimmer 30B, Muse Spark 1.2, and Muse Spark 1.1. Other listed Meta options include Llama Guard 4 12B, Llama 4 Maverick, and Llama 4 Scout. The facts here do not provide comparable specifications or prices for these alternatives.

Verdict

Muse Spark 1.3 Contributor has a clear stated niche: lower-cost experimentation and learning around multimodal reasoning, early-stage agents, multi-agent work, and coding. Its broad input types, tool and reasoning support, and large context window are useful facts for an initial evaluation. Its listing does not establish how those capabilities perform or how developers access them, so fit depends on confirming the integration details and testing the model against the needs of a particular project.