← Back to blog
Muse Ember Review: Official Size, Pricing and Access

Muse Ember Review: Official Size, Pricing and Access

Olivia Hart

By Olivia Hart

MyClaw Editorial

MyClaw

Run Best-in-Class AI Agents Now

Run OpenClaw or Hermes Agent on managed MyClaw hosting, so your AI agent stays online, updated, and ready for real work.

AI Takeaway

  • What is Muse Ember? Meta’s smallest Muse-family multimodal reasoning model for high-volume agent work: tool calls, coding subtasks, routing, extraction, and long documents. The API ID is muse-ember-1.0. Meta introduced it on 9 November 2026. It is not the Muse consumer agent, Muse Spark, or Muse Wonder.
  • How large is Muse Ember? Official docs list a 1,048,576-token context window and 524,288 maximum output tokens. It accepts text, image, video, audio, and PDF input, and returns text. Meta has not published a parameter count, public weights, or local hardware requirements.
  • What does Muse Ember cost on the official price list? On Meta Model API Standard rates, Muse Ember is $0.10 per million input tokens, $0.004 cached, and $0.50 output for requests up to 200K input tokens. When input exceeds 200K, the rates are $0.30 / $0.012 / $1.50, billed on the full request. There is no Contributor mode for Ember on the Model API. It is also on Muse Code pay-as-you-go and every subscription plan.
  • Who can use Muse Ember today? As of 11 November 2026, Meta says muse-ember-1.0 is available on the Meta Model API and in Muse Code. The launch post also names Google Cloud and Oracle Cloud. Official Ember materials do not label it a preview. Availability can still vary by jurisdiction.
  • Can you run Muse Ember on MyClaw? Not confirmed. Muse Ember is not listed on MyClaw. Use MyClaw Chat for everyday writing, research, and coding with models that are actually available.

What Muse Ember Is (and Is Not)

This Muse Ember review looks at the official size, pricing, and access for Meta’s smallest Muse-family multimodal reasoning model. Meta introduced Ember on 9 November 2026 for the work agents repeat: tool calls, coding subtasks, routing, extraction, and long documents. The Model API identifier is muse-ember-1.0.

Ember is not the Muse AI consumer agent, and it is not Muse Spark or Muse Wonder. Official docs put it in a supporting role: a larger model plans, and Ember handles the volume. Meta also says Ember is the first model trained using Muse Wonder Full.

The launch post positions Wonder for the hardest problems, Spark for coding and agentic workflows, and Ember for repeatable coding and agentic tasks at scale. On the model page, the same ID is described as a multimodal model for high-volume agentic workflow execution, including document summarization and extraction.

ItemOfficial detail (as of 11 November 2026)
API IDmuse-ember-1.0 (Standard mode)
Launch9 November 2026
Context window1,048,576 tokens
Max output524,288 tokens
Input / outputText, image, video, audio, and PDF in; text out
Reasoning effort"minimal" to "max"; default "max"
SurfacesMeta Model API, Muse Code, plus Google Cloud and Oracle Cloud in the launch post

Official Muse Ember Size and Specs

Official Muse Ember size, as listed in Meta’s docs, is a 1,048,576-token context window and 524,288 maximum output tokens. The launch post describes the same window as 1M tokens and the output cap as up to 512K tokens. Use the exact docs figures when you plan a request.

Inputs can be text, image, video, audio, and PDF; output is text. The launch post says Ember natively understands text, images, audio, and video, and the docs also list PDF. Reasoning effort runs from "minimal" to "max" and defaults to "max". Official docs list three endpoints: Responses, Chat Completions, and Messages.

Meta Launches Muse Ember, a Low-Cost AI Model

Meta has not published a parameter count, public weights, or local hardware requirements. Do not treat unofficial estimates as official size. The launch post says the long context window can take in a full codebase or document set in a single call; that is Meta’s wording, not a test in this review.

Official docs also list capabilities that this review has not tested: tool calling, structured output, computer use, and search grounding. For hard single-shot coding problems, knowledge-heavy questions, or visual coding, the same docs point to Muse Spark or Muse Wonder instead of Ember.

Official Muse Ember Pricing and Access

Official Muse Ember pricing on the Meta Model API is Standard-mode only; there is no Contributor mode for Ember on the Model API. List rates are billed per million tokens. The input-token count of the whole request, including cached tokens, sets the tier. A request with more than 200K input tokens is billed at the long-context rates in full, not only for tokens past 200K. Conversation history passed through previous_response_id counts as input.

Usage (per 1M tokens)Up to 200K inputOver 200K input
Cached input$0.004$0.012
Input$0.10$0.30
Output$0.50$1.50

Web-search grounding is $2.50 per 1,000 web_search queries, on top of token cost. Rate limits for muse-ember-1.0 are 3,000 RPM and 4,000,000 TPM per team.

For Spark’s Standard and Contributor prices, see the Muse Spark 1.3 review. As of 11 November 2026, official Standard rates are $1.25 input, $0.15 cached, and $4.25 output per million tokens for Spark, and $4.00 input, $0.20 cached, and $20.00 output for Muse Wonder. Spark and Wonder do not add a long-context premium on the published list; Ember does when input exceeds 200K tokens.

The launch post also cites vendor figures published on 9 November 2026, including “about $0.14 per Artificial Analysis task” and a claim of more intelligence per dollar than Claude Haiku 5.5 or GPT-6 Luna. Those are Meta and Artificial Analysis claims, not MyClaw tests.

Muse Ember - API Pricing & Benchmarks | OpenRouter

As of 11 November 2026, Meta says muse-ember-1.0 is available as pay-as-you-go Standard mode on the Meta Model API, and on Muse Code pay-as-you-go plus every subscription plan: Everyday at $5, High at $15, Power at $50, and Maximum at $200 per month. Official Ember materials do not label the model a preview. The launch post also names Google Cloud and Oracle Cloud. Availability can vary by jurisdiction.

On Model API Standard mode, Meta’s wording is: “In Standard mode, you pay standard prices and your content is not used to train Meta models.” On a Muse Code subscription, a request with more than 200K input tokens uses the plan’s usage faster. Recheck live rates and access before you budget.

Skip the Setup, Run Best-in-Class AI Agents Now

Launch a managed OpenClaw or Hermes Agent workspace in minutes, with always-on hosting, updates, and support handled by MyClaw.

Get Started

When to Use Muse Ember — and When Official Docs Point Elsewhere

Official choose-a-model guidance is Ember-first for sub-agent tasks, tool loops, routing, and the same job run over large collections of files or records. Use Spark or Wonder for the main loop and for harder single-shot work. Readers who need the frontier sibling can start with the Muse Wonder review rather than a full recap here.

The snapshot below comes from Meta’s launch table of Muse Ember (max) versus Claude Haiku 5.5 (max) and GPT-6 Luna (max). These are Meta and Artificial Analysis figures published on 9 November 2026, not independent MyClaw tests. FrontierCode 2.0-AA is a loss for Ember at 27.2 against Haiku 5.5 at 34.9, so the published set is not a sweep.

EvalEmber (max)Haiku 5.5 (max)Luna (max)
AutomationBench-AA37.813.920.2
Agents' Last Exam61.657.057.1
Terminal-Bench 4.040.939.216.4
Coding Agent Index-AA463731
FrontierCode 2.0-AA27.234.925.2
Artificial Analysis Index363628

What to Use on MyClaw If Muse Ember Is Not Listed

Muse Ember is not listed and not confirmed on MyClaw. Do not treat Ember, Spark, or Wonder as models you can run there.

For everyday writing, research, document work, and coding help, use MyClaw Chat with models that are actually available. On the chat page, describe a goal, add files or research if needed, choose a model or let routing pick one, and iterate in the same workspace. Available tools can depend on your account and plan. Do not treat Chat as a way to reach muse-ember-1.0.

Get Started

Conclusion

This Muse Ember review covers Meta’s smallest Muse model for high-volume agent execution, officially available as muse-ember-1.0 from 9 November 2026 on the Meta Model API and Muse Code. Official docs list a 1,048,576-token context window and 524,288 maximum output tokens. Model API Standard list rates are $0.10 input, $0.004 cached, and $0.50 output per million tokens up to 200K input, and $0.30 / $0.012 / $1.50 when input exceeds 200K, billed on the full request. Recheck live rates and access before budgeting. Ember is not the Muse consumer agent, not Spark, and not Wonder. Muse Ember is not listed or confirmed on MyClaw.

Skip the Setup, Run Best-in-Class AI Agents Now

Launch a managed OpenClaw or Hermes Agent workspace in minutes, with always-on hosting, updates, and support handled by MyClaw.

Muse Ember Review: Official Size, Pricing and Access | MyClaw.ai