Meta releases Muse Ember, a low-cost model for AI agent execution
Meta Superintelligence Lab released Muse Ember on November 9, 2026, presenting it as the lightest and most cost-efficient model in its Muse line. Meta aims it at agent workloads such as tool calls, coding subtasks, routing, extraction and long-document parsing, and prices it at 10 cents per million input tokens. Meta's cost-per-task comparisons, built on Artificial Analysis's composite benchmark, put Ember ahead of Anthropic's Haiku and OpenAI's Luna on cost and quality, and level with open-weight models such as DeepSeek V4.1 Flash and GLM 5.3 Flash on cost. The intended design pairs Ember with the larger Muse Wonder, which sets the rules and hands Ember the high-volume work, and Ember accepts text, image and audio inputs. This is product and industry news with no vulnerability, exploitation or threat activity. Meta's own announcement and the benchmark pages could not be retrieved during research, so these figures rest on the TechCrunch report and are unverified.
Linha do tempo
-
Meta releases Muse Ember, a low-cost model for AI agent execution
ler a análise desta fonte ocultar
Meta Superintelligence Lab released Muse Ember on November 9, 2026, presenting it as the lightest and most cost-efficient model in its Muse line. Meta aims it at agent workloads such as tool calls, coding subtasks, routing, extraction and long-document parsing, and prices it at 10 cents per million input tokens. Meta's cost-per-task comparisons, built on Artificial Analysis's composite benchmark, put Ember ahead of Anthropic's Haiku and OpenAI's Luna on cost and quality, and level with open-weight models such as DeepSeek V4.1 Flash and GLM 5.3 Flash on cost. The intended design pairs Ember with the larger Muse Wonder, which sets the rules and hands Ember the high-volume work, and Ember accepts text, image and audio inputs. This is product and industry news with no vulnerability, exploitation or threat activity. Meta's own announcement and the benchmark pages could not be retrieved during research, so these figures rest on the TechCrunch report and are unverified.