@codenameposhan: ElevenLabs just launched Eleven Realtime (October 20, 2026): two speech-to-speech models for building voice agents. v1 Mini is for natural conversation, v1 Max is for complex tasks. Here's what you need to know. 1. One system: speech recognition, turn-taking, the LLM and the voice are tuned together, end to end. It's available in ElevenAgents and through a Realtime API. 2. v1 Mini: 620 ms to first audio, 95.8% on Full-Duplex-Bench and a 988 Elo in the Speech Agent Arena. Built for assistants, tutors and characters, and it can tell when you're done talking or just saying "mm-hmm". 3. v1 Max: 570 ms to first audio, 98.8% on Big Bench Audio and an 87.1% task success rate. Built for customer support, lead qualification and scheduling, with tool calls mid-conversation. 4. Developers: the Realtime API is OpenAI-compatible. Change the endpoint and model name and keep the rest of your code, the same events and handlers. 5. Scale: 17,500+ voices, 90+ languages, SOC 2, HIPAA and GDPR support, EU data residency and zero retention. The catch: Mini and Max are shown on different benchmarks, every score comes from one evaluator (Artificial Analysis), and pricing is per minute plus LLM tokens, with Max costing more. Rates are on ElevenLabs' pricing page. Follow @codenameposhan for more AI breakdowns. Sources: ElevenLabs Eleven Realtime launch page · The Decoder, Oct 20, 2026 Independent briefing by CodenamePoshan, not an ElevenLabs post. The ElevenLabs logo and launch art belong to ElevenLabs, shown for editorial reporting. Platform logos are trademarks of their owners. #ElevenLabs #ElevenRealtime #VoiceAI #VoiceAgents #SpeechToSpeech #AI

Codename Poshan

Open In TikTok:

Region: US

Tuesday 20 October 2026 22:02:02 GMT