Meta logo

Muse Ember

Multimodal
Meta

Muse Ember is Meta Superintelligence Labs' smallest and most efficient multimodal reasoning model, built for high-volume agent execution such as tool calls, coding subtasks, routing, extraction and long-document work. It takes text, image, video, audio and PDF input and returns text, with a 1,048,576-token context window and up to 524,288 output tokens; reasoning effort runs from minimal to max with max as the default. On the Meta Model API it costs $0.10/$0.50 per million input/output tokens on requests up to 200K input tokens ($0.30/$1.50 beyond that), with cache reads at $0.004 and web-search grounding billed per 1,000 queries.

Key Specifications

Parameters
-
Context
1.0M
Release Date
November 9, 2026
Average Score
55.5%

Timeline

Key dates in the model's history
Announcement
November 9, 2026
Last Update
November 12, 2026
Today
November 20, 2026

Technical Specifications

Parameters
-
Training Tokens
-
Knowledge Cutoff
-
Family
-
Capabilities
MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)
$0.10
Output (per 1M tokens)
$0.50
Max Input Tokens
1.0M
Max Output Tokens
524.3K
Supported Features
Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Other Tests

Specialized benchmarks
Agents' Last Exam
Meta Muse Ember launch table, November 9, 2026; max reasoning effort. Official launch post. • Self-reported
61.6%
OSWorld 2.1 (partial)
Meta Muse Ember launch table, November 9, 2026; max reasoning effort. OSWorld 2.1 mean per-task partial-credit reward. Official launch post. • Self-reported
69.5%
OSWorld 2.1 (strict)
Meta Muse Ember launch table, November 9, 2026; max reasoning effort. OSWorld 2.1 strict binary completion secondary metric. Official launch post. • Self-reported
33.8%
PerceptionBench
Meta Muse Ember launch table, November 9, 2026; max reasoning effort. PerceptionBench scored with tools. Official launch post. • Self-reported
80.6%
Personal Superintelligence Index
Meta Muse Ember launch table, November 9, 2026; max reasoning effort. Internal composite of Agents' Last Exam, AutomationBench, FocusBench, JobBench, OfficeQA Pro, GDP.pdf, GDP.xlsx, and VBT. Reported on a 0-100 scale, stored as a fraction. Official launch post. • Self-reported
46.8%
Terminal-Bench 4.0
Meta Muse Ember launch table, November 9, 2026; max reasoning effort. Terminal-Bench 4.0 with the best published harness for this model (mini-swe-agent). Official launch post. • Self-reported
40.9%

License & Metadata

License
proprietary
Announcement Date
November 9, 2026
Last Updated
November 12, 2026

Similar Models

All Models

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.