Ember costs $0.10 per million input tokens and $0.50 per million output tokens. Direct rivals, Anthropic's Claude Haiku 5.5 and OpenAI's GPT-6 Luna, have identical list prices. Meta asserts Ember offers better performance per dollar than both rivals.
Mark Zuckerberg announced the launch on X. He called it a significant achievement. Ember excels in token-intensive daily tasks. Access is immediate via the Meta Model API and Muse Code. An independent evaluation by Artificial Analysis confirms Meta's claims. Ember scored 36 at its highest setting, matching Haiku 5.5. Luna scored 28. Using the index, Ember costs about $0.14 per task. Haiku was $0.41, and Luna was $0.15. Ember's cost is roughly one-third of Haiku's. Alexandr Wang stated Ember reached the „frontier of efficiency.” Agent and Coding Superiority Meta tests show Ember's strength in agent and coding tasks. In AutomationBench, Ember achieved 37.8%. This surpassed Luna by 20.2% and Haiku by 13.9%. However, on FrontierCode 2.0, Haiku scored 34.9%, while Ember scored 27.2%.
Some results stem from internal Meta evaluations. Ember works with larger models. It handles text, images, audio, and video inputs. Its memory capacity is impressive, analyzing up to one million tokens simultaneously. This handles entire codebases. Meta suggests pairing Ember with Muse Wonder. Wonder handles planning; Ember manages volume.
Ember was trained using the larger Muse Wonder Full model. Ember's weights remain proprietary, like other Muse series releases. Meta previously released a less powerful open-source model in October. Ember's entry price is competitive against Chinese rivals. These include DeepSeek V4.1 Flash, Z.ai's GLM-5.3 Flash, and MiniMax M3.1 Flash. These Asian competitors scored lower in the Artificial Analysis index. Meta applied internal safety frameworks. The model remained below high-risk thresholds for chemical, biological, and cyber threats.
Muse Ember is now available through the Meta Model API and Muse Code. It joins the Muse ecosystem powering Meta Muse. Meta is developing even more efficient models. These innovations are redefining AI market standards.