Muse Ember

At ten cents per million input tokens, Muse Ember is the cheapest route into Meta's Muse family that we list. Meta released it on 9 November 2026 and calls it its smallest and most efficient model, with the API identifier muse-ember-1.0. Output costs $0.50 per million tokens and cached input $0.004.

## The price step at 200,000 tokens

Rates triple once a prompt goes above 200,000 tokens, to $0.30 for input, $0.012 for cached input and $1.50 for output. Anyone planning to use the full 1,000,000-token context should cost it at the higher tier. Output can run to 512,000 tokens, a large allowance for a model at this price, and the Responses API is supported.

## Trained on a larger sibling

Meta says Ember is the first model trained using Muse Wonder Full. It claims more intelligence per dollar than Claude Haiku 5.5 or GPT-6 Luna, a comparison that comes from the vendor. The model is available through Muse Code and the Meta Model API, and Meta's page says Ember, Spark and Wonder can now also be reached on Google Cloud and Oracle Cloud. It states no open weights, and Artificial Analysis lists the model as proprietary.

## Independent figures so far

Artificial Analysis gives Muse Ember (Max) 36 on its Intelligence Index, against a median of 24 for comparable models, at $0.14 per index task and 3.1 minutes per task when read on 10 November 2026. Meta's own page puts Muse Spark 1.3 at 37 with prices of $1.25 and $4.25, so on Meta's figures Ember reaches nearly the same index score for a much smaller outlay.

## Key facts

**Maker:** Meta **Released:** 9 November 2026 **Model code:** muse-ember-1.0 **Context window:** 1,000,000 tokens **Maximum output:** 512,000 tokens **Input:** text and image, with audio and video stated by Meta **Price per million tokens up to 200K prompt tokens:** $0.10 input, $0.50 output **Availability:** Muse Code, Meta Model API, Google Cloud, Oracle Cloud

Not yet independently verified: Meta says Muse Ember understands audio and video natively. Artificial Analysis lists text and image input only, so audio and video have not been independently confirmed. We will update this when it can be confirmed, and remove this note.