Groq
Login

Muse Spark Open

PreviewEnterprise
meta-models/muse-spark-open
Try it in Playground
TOKEN SPEED
~350+ tps
Powered bygroq
INPUT
Text, images
OUTPUT
Text
CAPABILITIES
Tool Use, JSON Object Mode, JSON Schema Mode, Reasoning, Vision
Meta logoMeta
Model card

Muse Spark Open is Meta's open-weight model, an 816-billion-parameter Mixture-of-Experts reasoning model with 42B parameters active per token. It always reasons before answering, with adjustable reasoning effort, and supports tool calling, structured outputs, and image input for agentic and coding workflows. On Groq, Muse Spark Open is in limited preview. Contact us to discuss access for your organization.

LIMITS

CONTEXT WINDOW
131,072
MAX OUTPUT TOKENS
131,072
MAX INPUT IMAGES
5

QUANTIZATION

This uses Groq's TruePoint Numerics, which reduces precision only in areas that don't affect accuracy, preserving quality while delivering significant speedup over traditional approaches. Learn more here.

Key Technical Specifications

Model Architecture

A Mixture-of-Experts model with 816B total parameters and 42B activated per token, accepting text and image input and producing text output. Served on Groq with a 131K-token context window and up to 131K output tokens. Released under the Muse Model License.

Reasoning and Capabilities

Muse Spark Open is built for reasoning-heavy and agentic work:
  • Reasoning is always on, with reasoning_effort set to low, medium, high, or xhigh (default xhigh)
  • Tool calling, including parallel tool calls
  • Structured outputs with JSON schema
  • Image input alongside text

Use Cases

Agentic Coding and Tool Use
Plans and carries out multi-step tasks, calling tools and verifying results across long sessions.
  • Coding agents and multi-file code changes
  • Parallel function calling and tool orchestration
  • Long-horizon task planning and execution
  • Structured outputs for downstream systems
Complex Reasoning
Adjustable reasoning effort lets one model handle everything from quick answers to hard multi-step problems.
  • Math, science, and logic problems
  • Analysis over long documents
  • Visual reasoning over images
  • Research and decision support

Best Practices

  • Sampling: Muse Spark Open is tuned to perform best at its default temperature and top_p, so you can leave both unset.
  • Reasoning effort: the default xhigh delivers the model's strongest reasoning. Choose low or medium for snappier responses in interactive apps.
  • Stop sequences: stop applies to the final answer, so you can use stop strings freely without cutting reasoning short.

Get Started with Muse Spark Open

Once your organization has access, try Muse Spark Open on Groq:

shell
pip install groq
Python
from groq import Groq
client = Groq()
completion = client.chat.completions.create(
    model="meta-models/muse-spark-open",
    messages=[
        {
            "role": "user",
            "content": "Explain why fast inference is critical for reasoning models"
        }
    ]
)
print(completion.choices[0].message.content)

Was this page helpful?