Google releases Gemini 3.8 Live speech model, tops Speech to Speech Index at 82.6, costs $0.84/hour
ArtificialAnlys · x · 2026-09-16
Google DeepMind released Gemini 3.8 Live, successor to Gemini 3.1 Flash Live — a Speech to Speech model that executes tools and API calls mid-conversation, in standard and Extended Thinking variants.
Artificial Analysis evaluation highlights:
- Index: Extended Thinking (High) debuts #1 at 82.6, ahead of GPT-Live-1 (Astra) at 81.5 and Grok Voice Think Fast 2.0 High at 81.3; the standard model is #5 at 76.0.
- Agentic: #1 on Tau Voice at 68.6% (up from 37.7% for the previous gen); #2 in Speech Agent Arena preference (Elo 1083) with 93.2% task success.
- Audio reasoning: 97.7% on Big Bench Audio, behind only Qwen Audio 3.0 Realtime Plus (99.2%).
- Speed: Time to First Audio of 1.18s (standard) / 1.35s (ET High), vs 2.99s for the previous gen; Grok remains fastest at 0.70s.
- Cost: $0.84/hour of input audio, cheapest on the index and roughly half the previous gen; ET High at $3.50/hour is 3.1x cheaper than GPT-Realtime-2.1 High ($10.75).
More from Models
- Fable 5.5 creates 66-second animated short entirely with self-written code, no image models — oran_ge · 2026-11-02
- SenseTime open-sources SenseNova-Loop-9B: weight-shared 56-layer loop, 256K context, Apache 2.0 — AdinaYakup · 2026-11-02
- User says Gemini excels at long-form fanfic, producing stories up to 29 minutes of reading — Ok-Jello3739 · 2026-11-02
- Sam Altman calls tokens 'a kind of fake unit' that doesn't measure business value — rohanpaul_ai · 2026-11-02
- 'You Can Feel the Gradient Slopes': Dev Contrasts Benchmaxxed RL Models vs General Ones — theteknosaur · 2026-11-02
- LightOnOCR-3 Launches: Tops FRBench, Adds Layout Structure and Chart Extraction — IgorCarron · 2026-11-02