~/ARTIFICIAL I/meta-launches-muse-ember-lightweight-ai-model-matching-claude-haiku-at-one

Meta Launches Muse Ember: Lightweight AI Model Matching Claude Haiku at One-Third Cost

Meta has introduced Muse Ember, its smallest and most efficient AI model to date, designed specifically for complex reasoning and routine agentic tasks. According to Artificial Analysis, it matches the performance of Claude Haiku 5.5 while costing only one-third per task. This release marks a significant milestone in cost-efficient AI inference, enabling developers and enterprises to run advanced multi-modal agents and handle massive codebases at a fraction of previous expenses. Its competitive pricing and high benchmark scores intensify the market competition for lightweight, high-performance language models. Muse Ember features a 1-million token context window, supports a maximum output of 51.2K tokens, and natively processes text, images, audio, and video. It is available immediately via Meta Model API, Google Cloud, and Oracle Cloud, priced at $0.10 per million input tokens and $0.50 per million output tokens.

## BACKGROUND

AI inference costs and context window limitations are critical factors for enterprises deploying large language models at scale for autonomous agent workflows. Lightweight models bridge the gap by offering high throughput and low latency for routine coding, routing, and tool-calling subtasks without requiring massive computational resources.

## REFERENCES

## KEYWORDS

#Artificial Intelligence#Meta#LLM#AI Models#Cost Efficiency

$ subscribe --daily