Models
Meta releases Muse Ember, its smallest model, at $0.10 and $0.50 per million tokens: the same list price as Claude Haiku 5.5 and GPT-6 Luna, with about a third of Haiku’s cost per task at an equal index score, per Meta’s chart
Meta’s Muse Ember lists at $0.10/$0.50 per million tokens like Haiku 5.5 and Luna; Meta says it costs about a third per task at an equal index score.
Meta released Muse Ember on November 9, 2026, its smallest and cheapest model, built for the repetitive work of AI agents. Its list price matches Anthropic’s Claude Haiku 5.5 and OpenAI’s GPT-6 Luna, so Meta’s claim of better value rests on tokens spent per task, not on the sticker price. The cost figures come from Meta’s own chart of Artificial Analysis data.
The essentials
- What launched: Muse Ember, API id muse-ember-1.0, is on the Meta Model API and in Muse Code. It takes text, images, audio and video, with a 1M-token context and up to 512K output tokens. The model is proprietary and Meta did not disclose its size.
- The price claim: $0.10 per million input tokens and $0.50 per million output, identical to Haiku 5.5 and Luna for short prompts. So “more cost-effective” means fewer tokens per task, not a lower list price.
- The numbers: At maximum reasoning, Ember scores about 36 on Artificial Analysis’s Intelligence Index, level with Haiku 5.5 (36.2) and above Luna (27.8). Meta’s chart shows roughly $0.14 per task for Ember, $0.41 for Haiku and $0.15 for Luna; Ember’s figure is an estimate at proposed pricing.
- Caveats: Pasquale Pillitteri says Ember trails Haiku on FrontierCode. Simon Willison’s tables show all three models’ prices rise above a prompt-size threshold. On the same chart, Muse Spark 1.3 scores 37 at $2.39 per task.
What “cheaper” means here
Two models with the same price per token can differ in cost per task if one writes more tokens to finish the job. Meta says Ember spends fewer, which is why it reports about a third of Haiku’s cost at the same index score. Whether that holds on your own workload is untested.
Limits of the comparison
The index is independent, but the cost-per-task chart and the estimate for Ember are Meta’s presentation. We read relays and extracts, not the benchmark data, and found no independent replication yet. Not investment advice.
What is not established
- Ember’s cost per task on real workloads: only Meta’s chart and an estimate at proposed pricing.
- Ember’s parameter count: not disclosed.
- Ember against Haiku 5.5 on coding: one relay says it trails on FrontierCode; not checked.
- Independent replication of the cost-per-task gap: none found.
Summary table
| Point | What is stated as of November 7, 2026 | Source |
|---|---|---|
| Release | November 9, 2026, on the Meta Model API and in Muse Code | How2Shout |
| List price | $0.10 input and $0.50 output per million tokens, same as Haiku 5.5 and Luna | The Next Web |
| Index score | About 36, level with Haiku 5.5 and above Luna’s 27.8 | Interesting Engineering |
| Cost per task | About $0.14 against $0.41 for Haiku 5.5, per Meta’s chart | The Next Web |
| Weak spot | Trails Haiku on FrontierCode, per one relay | Pasquale Pillitteri |
FAQ
Is Muse Ember cheaper per token than Haiku 5.5?
No. Short-prompt list prices are identical; Meta’s claim is about fewer tokens spent per task.
Who measured the score?
Artificial Analysis produced the index; the cost-per-task chart and Ember’s estimate are presented by Meta.
Is it open?
No. It is a proprietary model available through Meta’s API and Muse Code.
Method and limits
Sources checked on November 10, 2026: The Next Web, Simon Willison, Interesting Engineering, Pasquale Pillitteri and How2Shout, read as search extracts; Meta’s announcement and the benchmark data were not read directly. Not investment advice. Disclosure: OrigoTech drafts with Claude, a model made by Anthropic, whose Haiku 5.5 is compared in this article.
Sources
- The Next Web: Meta’s Muse Ember matches Claude Haiku at a third of the cost per task (verified 2026-11-08)
- Simon Willison: Introducing Our Smallest, Most Efficient Model: Muse Ember (verified 2026-11-08)
- Interesting Engineering: Muse Ember: Meta unveils its smallest model yet with 1M-token context for cheaper agents (verified 2026-11-08)
- Pasquale Pillitteri: Meta launches Muse Ember, its cheapest AI agent model at $0.10 per million tokens (verified 2026-11-08)
- How2Shout: Meta Unveils Muse Ember: Lightweight AI That Matches Claude Haiku 5.5 at One-Third the Cost (verified 2026-11-08)
Spotted an error in this article? Report it · Our corrections policy