The race to build more capable artificial intelligence models is moving into a new phase, with Meta and Anthropic unveiling advances aimed at making AI systems more useful for complex, sustained work. The announcements highlight a competitive push to expand what AI can accomplish, from coordinating multiple agents to handling sophisticated software engineering and other technical tasks.
Meta’s Muse Ember, unveiled Monday (Nov. 9), is described as a smaller, more efficient model built to handle repetitive tasks performed by AI agents, including tool calls, coding subtasks, information extraction and document analysis. It can also coordinate business workflows across applications and process large collections of documents or code within a single request.
The model is available through Meta’s Model API at $0.10 per million input tokens and $0.50 per million output tokens. Meta said its internal and third-party evaluations show Muse Ember performing competitively against models from Anthropic and OpenAI, particularly in agent-based workflows and coding. The company also highlighted its ability to process text, images, audio and video, positioning the model for a broad range of enterprise applications.
Anthropic’s Fable 5.2 and Mythos 5.2, also unveiled Monday, are aimed at more demanding technical assignments. Fable can coordinate more than 1,000 AI agents simultaneously, tackle advanced software engineering projects and handle hardware-related work, including controlling robots. Anthropic said both models are faster than their predecessors, while Fable is less expensive.
The company reduced Fable’s pricing to $8 per million input tokens and $40 per million output tokens, down from $10 and $50, respectively, for the previous version. Mythos 5.2 uses the same underlying model but has fewer safeguards for cybersecurity and life sciences applications. Access remains restricted to organizations participating in Anthropic’s verification program.
The releases also bring renewed attention to the safety challenges accompanying increasingly capable AI systems. Anthropic said it subjected the new models to longer testing periods and additional evaluations, including assessments of their ability to escape digital sandbox environments. The company said the latest versions were less likely to breach those restrictions than their predecessors.
For both companies, the announcements underscore a market focused on making AI agents practical for business use. Greater reasoning capabilities remain important, but cost, speed and the ability to execute complex workflows are becoming equally significant measures of competitiveness.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
