This applies to people who call models from their own code or run coding agents such as Claude Code, Codex, or Cursor through AI Gateway. It does not affect v0 editor users or people who build sites without code, since the announcement does not mention v0.
What is new
- Muse Ember from Meta can be called on AI Gateway with the model name meta/muse-ember.
- The changelog says it accepts text, image, and video input and has a 1M-token context window.
- It supports tool calling and reasoning effort that you can set from minimal to max.
- You can reach it through the AI SDK, Chat Completions, Responses, the Anthropic Messages API, or a coding agent connected to AI Gateway.
- The model page lists $0.10 per million input tokens and $0.50 per million output tokens.
Taken from the official update: vercel.com [1]
What Muse Ember is meant to do
Vercel presents Muse Ember as suited to bounded tasks that an agent repeats many times in a run. Examples include implementing a helper function, making a tool call, or summarizing a document. Tool calling allows your application to pass each result into the next step.
Fact checked on 13 Nov 2026 against vercel.com [1]
The model page lists a context window of 1,048,576 tokens, shared between text prompts and responses. Output can also be up to 1,048,576 tokens, with reasoning tokens included in that limit. Increasing reasoning effort leaves less room for the visible answer in a very long response.
Fact checked on 13 Nov 2026 against vercel.com [2], vercel.com [1]
Vercel's pages differ on one detail. The changelog says the model accepts text, image, and video input. The model page lists text, image, and PDF as input types, but not video. Check the model page before building anything that relies on video.
Fact checked on 13 Nov 2026 against vercel.com [1], vercel.com [2]
How you call it
In the AI SDK, the model can be specified in one line, for example model: 'meta/muse-ember' within a streamText call. It is also listed for Chat Completions, Responses, and Anthropic Messages calls, allowing existing code written for those formats to use it.
Fact checked on 13 Nov 2026 against vercel.com [1]
For coding agents, Vercel instructs users to install the latest Vercel CLI, run npx vercel ai-gateway setup, and select the model in the agent. Vercel also provides a link to a model playground for trying it first.
Fact checked on 13 Nov 2026 against vercel.com [1]
What it costs
The model page lists prices of $0.10 per million input tokens, $0.50 per million output tokens, and $0.004 per million cache-read tokens. Vercel says AI Gateway charges the provider's list price without a markup or platform fee on tokens, including requests made with your own provider key.
Fact checked on 13 Nov 2026 against vercel.com [2], vercel.com [3]
Here is an example calculation using the listed prices, not a measured result. Suppose an agent makes 10,000 small calls, each using 2,000 input tokens and 500 output tokens. That comes to 20 million input tokens ($2.00) plus 5 million output tokens ($2.50), for a total of $4.50. Reasoning tokens count toward output, so higher reasoning effort can increase the output portion of the bill.
Fact checked on 13 Nov 2026 against vercel.com [2]
Using the same calculation with the starting prices of other lightweight models listed on AI Gateway gives about $12.50 for Gemini 3.1 Flash Lite ($0.25 input, $1.50 output) and about $37.50 for GPT 5.4 Mini ($0.75 input, $4.50 output). Claude Haiku 5.5 is listed at the same $0.10 and $0.50 as Muse Ember. Prices on those pages may change, so check them before comparing.
Fact checked on 13 Nov 2026 against vercel.com [4], vercel.com [5], vercel.com [6]
AI Gateway offers a free tier for a subset of models and a paid tier for all available models. It does not specify whether Muse Ember is included in the free subset.
Fact checked on 13 Nov 2026 against vercel.com [3]
What the official notes do not say
- The official pages do not say whether Muse Ember is eligible for the free AI Gateway tier.
- The update does not say which model or feature Muse Ember replaces.
- The pages do not give a Muse Ember rollout schedule or any extra plan or version requirement.
- The pages do not give a rate limit specific to Muse Ember.
- The changelog lists video input but the model page does not, so check the model page for video support before relying on it.
- The announcement does not mention v0, so it does not say whether the v0 editor offers this model.
How it helps: a freelancer who builds sites for local shops and has a coding agent write many small form-validation helpers
The task. Have the agent create a batch of small helpers, such as checks for phone number and email fields, and summarize a client’s long brief along with its screenshots in one run using a low-cost model.
- Install the latest Vercel CLI and run npx vercel ai-gateway setup.
- Open your coding agent, such as Claude Code, Codex, or Cursor, and select meta/muse-ember as the model.
- Set a low reasoning effort for simple helpers and a higher one for harder tasks, anywhere from minimal to max.
- Hand over the client's brief and screenshots as input, after checking the model page for the input types it lists.
- Estimate the bill with the listed prices before a long job. For example, 10,000 calls of 2,000 input and 500 output tokens come to $4.50.
The result. The repeated subtasks use the same gateway as your other models, at $0.10 per million input tokens and $0.50 per million output tokens. This post makes no claims about speed or quality.
Fact checked on 13 Nov 2026 against vercel.com [1], vercel.com [2]
Other ways to use it
Muse Ember accepts image input and has a 1M-token window, allowing one run to take a long brief along with its screenshots. Long inputs are charged per input token, so check the pricing first.
Fact checked on 13 Nov 2026 against vercel.com [1], vercel.com [2]
To switch, change the model string to meta/muse-ember. Vercel says the gateway adds no markup to token prices.
Fact checked on 13 Nov 2026 against vercel.com [1], vercel.com [3]
Sources
- https://vercel.com/changelog/muse-ember-now-available-on-ai-gateway (official update)
- https://vercel.com/ai-gateway/models/muse-ember
- https://vercel.com/docs/ai-gateway/pricing
- https://vercel.com/ai-gateway/models/gemini-3.1-flash-lite
- https://vercel.com/ai-gateway/models/gpt-5.4-mini
- https://vercel.com/ai-gateway/models/claude-haiku-5.5
Facts in this article were checked against these sources on 13 Nov 2026. Product details change, so check the official page before you rely on them.