Meta Superintelligence Lab released a new AI model on Monday called Muse Ember, described as the lab's lightest and most cost-efficient model yet. According to TechCrunch, the model is designed for complex reasoning execution and the repetitive work agents handle all day: tool calls, coding subtasks, routing, extraction, and long document parsing.
The release puts pressure on rivals on price. At 10 cents per million input tokens, Muse Ember is significantly cheaper than frontier competitors such as Anthropic's Haiku and OpenAI's Luna, and it is cost-competitive with open-source options like DeepSeek V4.1 Flash and GLM 5.3 Flash, Meta said.
Muse Ember specs and pricing
Meta particularly emphasized cost-per-task, using Artificial Analysis's composite benchmark score to show Ember as cheaper and smarter than competitors like Haiku and Luna, according to TechCrunch. The company also trained the model to handle several types of media, including text, images, and audio.
That mix matters for agents that must work across messy real-world material. As Meta puts it in the announcement, agents can read screenshots, charts, and scanned documents, or work from voice memos, meeting recordings, and video without a separate model.
- Price: 10 cents per million input tokens
- Built for: tool calls, coding subtasks, routing, extraction, long document parsing
- Inputs: text, images, and audio without a separate model
- Benchmark: Artificial Analysis composite score, positioned as cheaper and smarter than Haiku and Luna
- Role: pairs with the larger Muse Wonder model in an agentic harness
