Google launched Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on September 15. The Extended Thinking model secured the top spot on the Artificial Analysis Speech-to-Speech Index with a score of 82.6, surpassing GPT-Live-1 and Grok Voice. The standard Gemini 3.8 Live variant ranked fifth with 76.0 points.
The models demonstrate improved performance in agentic tasks and speech reasoning. Extended Thinking achieved 68.6% on the Tau Voice benchmark and 97.7% in speech reasoning. Latency for the new models decreased significantly, with the average time to first audio dropping to 1.18 seconds from 2.99 seconds in the previous generation.
Pricing for the new models varies by version. The standard Gemini 3.8 Live costs $0.84 per hour of input audio, while the Extended Thinking version is priced at $3.50 per hour. These rates are lower than competing models like GPT-Live-1 Sol at $4.47 and Grok Voice Think Fast 2.0 High at $4.80.
Despite high benchmark scores, user preference remains with older technology. Gemini 3.1 Flash Live currently holds the highest Elo rating of 1096 in blind Speech Agent Arena tests, followed by Gemini 3.8 Live at 1083. The Extended Thinking variant holds an Elo of 990, despite a task completion rate of 89.1%.
