Key points
- Meta introduced Muse Ember as an entry-level artificial intelligence model.
- The company claims it costs significantly less than competing models.
- Benchmarks show it matches Anthropic’s Claude Haiku 5.5 at lower prices.
SiliconANGLE reported that the entry-level model costs significantly less than the competition.reported
The evidence
Meta states that Muse Ember is the first AI model trained using Muse Wonder. Muse Wonder is the company’s flagship large language model. The source material does not specify which part of the workflow Meta automated with Muse Wonder during training.
Meta evaluated Muse Ember using the Artificial Analysis Index. This index aggregates nine AI benchmarks covering use cases such as coding and financial analysis. According to Meta, Muse Ember matched the score of Anthropic’s entry-level Claude Haiku 5.5 model at about a third of the price.reported
Meta also reported that Muse Ember outperformed OpenAI’s comparable GPT-6 Luna algorithm in the same evaluation.reported In a separate internal test, Meta compared Muse Ember against Haiku and Luna across nine other benchmarks. All three algorithms were set to max mode, using maximum hardware resources.
The decision
In the max mode tests, Muse Ember outperformed both Haiku and Luna across seven of the nine benchmarks. The area where the model appears to have the biggest edge is business task automation.
On the AutomationBench-AA benchmark, Muse Ember recorded a 17% lead over the rivals.reported This benchmark collection spans 657 tasks across popular cloud applications such as Gmail and Salesforce. Most other benchmark results showed only a few percentage points of difference between the models.
What to make of it
Meta says Muse Ember is specifically optimized for tasks assigned to agents in AI ensembles. Applications often use frontier LLMs like Muse Wonder to split user requests into smaller tasks. These smaller tasks are then assigned to agents powered by a lower-cost model.
SiliconANGLE noted that Thinking Machines Labs recently demonstrated small LLMs can run post-training end-to-end. Anthropic also revealed its AI models could soon accelerate machine learning research. Meta did not confirm if it used Muse Wonder for post-training.
In practice
Users can access Muse Ember with a context window of 1 million tokens.reported This allows the model to ingest lengthy business documents. Users can also include images, audio and video in their prompts.
The model generates responses up to 512,000 tokens in length.reported
Why it matters
- Founders can potentially reduce inference costs by using Muse Ember for sub-tasks in AI ensembles instead of more expensive frontier models.
- The 1 million token context window allows processing of long business documents, which is relevant for automated analysis workflows.
- Meta’s claim of matching Haiku 5.5 at one-third the price offers a direct cost-benefit comparison for budget-conscious development teams.
- The 17% lead on AutomationBench-AA suggests stronger performance for cloud application tasks like Gmail and Salesforce automation.
Topics: models, Meta, AI agents, benchmarking, cost
Updated
For readers
Developers
- Muse Ember has a 1 million token context window and supports images, audio and video in prompts.
- The model generates responses up to 512,000 tokens in length.
Business
- Meta claims Muse Ember matches Haiku 5.5 scores at about a third of the price.
- The model showed a 17% lead on business task automation benchmarks like AutomationBench-AA.
Investors
- Muse Ember is optimised for lower-cost model roles in AI ensembles.
- Meta introduced Muse Ember as a cost-efficient option compared to competition.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
Muse Ember
An entry-level artificial intelligence model introduced by Meta Platforms Inc.
Muse Wonder
Meta Platforms Inc.'s flagship large language model used to train Muse Ember.
Artificial Analysis Index
An index aggregating nine AI benchmarks covering use cases such as coding and financial analysis.
AutomationBench-AA
A collection of 657 tasks spanning popular cloud applications such as Gmail and Salesforce.
Questions readers ask
How does Muse Ember compare to competitors?
Meta says Muse Ember matched Anthropic’s Claude Haiku 5.5 score at about a third of the price and outperformed OpenAI’s GPT-6 Luna.
What is the context window size?
Muse Ember has a context window of 1 million tokens and can generate responses up to 512,000 tokens in length.
Connected stories
Reading receipt
- Meta introduced Muse Ember as an entry-level artificial intelligence model.
- The company claims it costs significantly less than competing models.
- Benchmarks show it matches Anthropic’s Claude Haiku 5.5 at lower prices.