Muse Spark 1.3
Language modelCard summarizedMuse Spark 1.3 is trained for agentic workflows and optimized for competitive coding performance.
● Re-checked 12 Oct 2026 · 72 facts · 2 corrections · 2 unconfirmedSource: Muse Spark 1.3 model page, launch post and evaluation methodology (no system card published)
Use it via API
Where to call it
- Meta Model API
muse-spark-1.3 - Google Vertex AI
—
Supports
- Image input
- PDF input
- Structured outputs
- Reasoning control
- Tool calling
- Web search
- Prompt caching
Good for
- Long-horizon agentic workflows: multi-step tool, browser and long-running tasks in one thread
- Agentic coding with fewer unneeded turns and less verbose output than 1.2
- Following long, complex instructions without dropping constraints
Overview
Muse Spark 1.3
Muse Spark 1.3 is trained for agentic workflows and optimized for competitive coding performance.
- Cost
- $1.25 in
- Context
- 1M
- Input / output
- $1.25 / $4.25
- Max output
- —
- Knows up to
- —
- Released
- 2 Sep 2026
Meta's current Muse Spark version, tuned for long-horizon agentic workflows and coding, and the first Spark version with a max reasoning level. Meta has published no system card for 1.3; this summary uses its model page, launch post and evaluation methodology. In Meta engineers' comparisons it used about 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2. launch post; model page
muse-spark-1.3- Input / output per 1M
- $1.25 / $4.25
- Context window
- 1M
- Max output
- —
- Knows up to
- —
- Reasoning
- Always on; reasoning_effort minimal to xhigh, plus max (Standard tier only)
- Weights
- Closed
About this card. A model card is the maker's own account of a model: what it's for, how it was trained and tested, and where it falls short. This page condenses Meta's Muse Spark 1.3 model page, launch post and evaluation methodology (no system card published). Every claim carries its source page.