Meta 发布 Muse Ember：多模态模型，智能体与编程基准超过 Claude Haiku 5.5 和 GPT-6 Luna
Meta 的新模型在编程和 agent 基准上超过了 Claude Haiku 5.5 和 GPT-6 Luna，成本只有 Haiku 的三分之一，还原生支持音频视频多模态。做智能体应用、想压成本的开发者可以直接对比试试。
Meta 发布新模型 Muse Ember，在关键智能体（agent）和编程基准上超过 Anthropic 的 Claude Haiku 5.
5 与 OpenAI 的 GPT-6 Luna。定价方面，三款模型均从每百万输入/输出 token 0.
10/0.50 美元起，而 Ember 的综合智能分数与 Haiku 5.
5 持平，单任务成本约为其三分之一。Ember 是 Meta 首个使用 Muse Wonder Full 训练的模型，支持文本、图像、音频和视频四种模态。
由于三家定价起点相同，Ember 的成本优势主要体现在单任务实际消耗上。对开发者而言，这意味着在智能体和编程场景下多了一个更低成本的选择。
First release of the day: Meta’s new Muse Ember beats Claude Haiku 5.5 and GPT-6 Luna on key agent and coding benchmarks.
It matches Haiku's overall intelligence score at roughly one-third the cost per task. All three start at $0.10/$0.50 per million input/output tokens.
Ember is Meta’s first model trained using Muse Wonder Full and supports text, images, audio and video.
Great launch and release! Meta is super back!
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Muse Ember, Meta, Claude Haiku, GPT-6.
Source attribution: curated from kimmonismus (@kimmonismus). Cross-source corroboration: 11 independent sources have covered this development (including Artificial Analysis, techcrunch, Meta AI), confirming industry-wide signal strength.
