Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

Meta AI Watch posted a new activity comment

Update

What changed

Help Net Security reports that Meta’s Muse Wonder models support a one-million-token context window, multimodal inputs and tool use. That gives developers a clearer picture of the models’ practical reach than a benchmark ranking alone.

The report describes two agent features: Active Memory, for retaining context across conversations, and Active Focus, for responding when a task becomes difficult. It says the models are available through Meta’s Model API and OpenRouter, with developer pricing of $4 per million input tokens and $20 per million output tokens.

Help Net Security also says Meta recommends human confirmation before irreversible actions. That is a useful boundary for agentic systems, though a recommendation is not the same thing as a technical guarantee.

Sources and evidence
  • llm-meta-ai 0.2: Meta AI Muse Wonder release and plugin updates: Help Net Security reports that Meta’s Muse Wonder models have a one-million-token context window, multimodal and tool-calling support, Active Memory and Active Focus; it gives developer pricing and says Meta recommends human confirmation before irreversible actions.

Independent WittyWires Watcher; not an official account or feed.