ElevenLabs released two voice agent models that handle listening, thinking and speaking in one piece, which could make phone and intake bots simpler to build.
ElevenLabs has launched Eleven Realtime, a pair of AI models that can run a spoken conversation on their own. Until now, building a voice agent that sounds natural meant stitching together several separate parts. This new class of models replaces that stack with one. The company calls them Speech to Speech models.
A typical voice agent needs speech recognition, a way to detect when someone stops talking, rules for taking turns, a language model to decide what to say, and a voice to say it. ElevenLabs says getting those parts to work well together is a project in itself. Each Eleven Realtime model is a complete voice agent. Both support more than 90 languages and can use any of the 17,000+ voices in the ElevenLabs library.
There are two models. Max is the stronger one. Mini is the lighter option. Artificial Analysis, an outside group, tested both. ElevenLabs reports these results:
In blind live conversations, where people compared two models, Max scored 998 and Mini 988. ElevenLabs recommends Max for agents that need to take actions, such as looking things up or updating records.
There are two routes. In ElevenAgents, the company's platform for running agents, you pick a Realtime model from a dropdown. That brings phone, web, WhatsApp and call center software channels, plus workflows, a knowledge base, guardrails, testing and monitoring. Settings such as how long to wait after a pause are adjustable.
The second route is the Realtime API, for developers. It follows OpenAI's Realtime API specs, so code written for that service may only need a new endpoint and model name.
The Realtime API connects over WebSocket only, so a setup built on WebRTC has to move first. It also runs separately from ElevenAgents. A business that wants channels or a knowledge base should build on ElevenAgents from the start. The announcement does not give prices, and the benchmark numbers come from tests, not from real customer calls.
Businesses that answer calls, take bookings or collect intake details by voice are the main audience. Fewer parts to tune could mean faster builds. The tool-use gap between the two models matters most for anyone planning an agent that does more than chat.
Eleven Realtime packs a whole voice agent into one model, and Max is the pick if the agent must take actions, not just talk.
Otto, a new customer-tracking app for field sales teams, lets reps speak their notes instead of typing them. Price and launch details are not yet known.
Read more: Otto launches a sales app you talk to instead of typingA new Alibaba-led test of real coding requests found the best AI agent fully finished just 26.4 percent of tasks, so human review still matters.
Read more: AI coding helpers fully finish only about 1 in 4 real tasksEA has re-released 25 classic PC games at a discount until November 19, and Bethesda has finally given The Elder Scrolls VI a subtitle: Sentinel.
Read more: EA brings 25 classic PC games back, and Elder Scrolls VI gets a name
