Meta is entering the high-performance AI model market with the release of Muse Ember, developed as a direct competitor to models from OpenAI (GPT-6 Luna) and Anthropic (Claude Haiku 5.5).
The release was preceded by a prolonged struggle in the affordable technology segment. Several months ago, Chinese laboratories released a line of models that delivered about 80–90% of the performance of leading solutions at a significantly lower price. OpenAI responded by launching GPT-6 Sol and GPT-6 Luna, and Anthropic last month presented Haiku 5.5 for processing large volumes of data. These steps curbed the growing popularity of Chinese alternatives in the US.
Now Meta has presented its smallest and most efficient model, Muse Ember, which is not inferior to competitors in intelligence level but costs significantly less, writes Neowin.
In the Artificial Analysis Intelligence Index v5 test, the Ember model scored 36 points, matching Claude Haiku 5.5 and surpassing GPT-6 Luna (28 points).
At the maximum reasoning level, one query in Muse Ember costs about $0.14, whereas Claude Haiku 5.5 costs $0.41 and GPT-6 Luna costs $0.15. This delivers Haiku-level quality at almost three times cheaper.
Claude Haiku 5.5 still retains leadership in the FrontierCode 2.0 benchmark (an assessment of real software engineering tasks), but Ember supports multimodality — input data in the form of text, images, audio, and video. The context window is 1 million tokens, and the maximum generation volume is up to 512,000 tokens.
The price of Muse Ember is $0.10 per million input tokens, $0.004 per million cached tokens, and $0.50 per million output tokens (for prompts up to 200,000 tokens). For longer requests, the price rises to $0.30 for input and $1.50 for output tokens.
The model is already available through the Meta Model API and Muse Code (ID: muse-ember-1.0). It supports the OpenAI Responses API specification, which allows developers to easily integrate it into existing agentic scenarios.