Raw1 AI – 100% Local AI.

Raw1 AI Model Release News – September 16, 2026

AI MODEL MONITOR

AI Model Release Monitor — Daily Digest (2026-09-16)

2026-09-16

1 new model release(s) found:

OTHER RELEASES

RSS/MarkTechPost
Two new native speech-to-speech voice-agent models built on Gemini 3 Pro: standard 3.8 Live is tuned for low-latency, cost-efficient conversation, while 3.8 Live Extended Thinking adds background multi-step reasoning and async tool calls while continuing to speak. Both handle live audio/video/image input across 97 languages.
Source: https://deepmind.google/blog/introducing-gemini-3-8-live-and-3-8-live-extended-thinking/
Hardware: Closed, hosted-only model served via the Gemini API/AI Studio (no open weights) – no self-hosted hardware requirement; billed per-minute ($0.005/min audio in, $0.018/min audio out).

Tracking 18 models total (14 China-origin) since monitoring began.

Full browsable, searchable model page
Source: https://drive.google.com/file/d/1NpL8_FQv2YM5fM2x8N9A5w8jgfi3xrjQ/view?usp=sharing