GoogleAudio matchupReal-time voice dialog
Gemini 3.8 Livevs.Gemini 3.8 Live Extended Thinking
Which model is better?
Gemini 3.8 Live Extended Thinking does better in independent tests.
Pick Gemini 3.8 Live for
Pick Gemini 3.8 Live Extended Thinking for
Shared tests
Independent evaluators and leaderboards only, the newest shared snapshot per evaluator. With 95 % ranges a lead counts when they do not overlap; for hit rates, when it holds in a 95 % test. Without any uncertainty figure at most “slightly ahead”.
Speko
depends on the subtestGemini 3.8 Live is clearly ahead in 0 of 7 subtests and slightly in 3, Gemini 3.8 Live Extended Thinking in 0 clearly and slightly in 3. 1 is level.
Speko S2S first-sound latency p50
lower is better, as of October 28, 2026
Speko S2S task-content latency p50
lower is better, as of October 28, 2026
Speko S2S booking before caller confirmation
lower is better, as of October 28, 2026
Show all 7 subtests
| Subtest | Gemini 3.8 Live | Gemini 3.8 Live Extended Thinking | Verdict |
|---|---|---|---|
| Speko S2S first-sound latency p50 (lower is better) | 1,242 ms | 863 ms | Gemini 3.8 Live Extended Thinking slightly ahead |
| Speko S2S task-content latency p50 (lower is better) | 1,569 ms | 1,246 ms | Gemini 3.8 Live Extended Thinking slightly ahead |
| Speko S2S booking before caller confirmation (lower is better) | 18.0 % | 22.0 % | Gemini 3.8 Live slightly ahead |
| Speko S2S task capability mean | 89.0 | 84.0 | Gemini 3.8 Live slightly ahead |
| Speko S2S confirmation-safe action success | 68.0 | 72.0 | Gemini 3.8 Live Extended Thinking slightly ahead |
| Speko S2S overall call quality | 79.0 | 75.0 | Gemini 3.8 Live slightly ahead |
| Speko S2S scripted call completion | 96.0 % | 94.0 % | level |
Source · latest October 28, 2026
Artificial Analysis
Gemini 3.8 Live Extended Thinking aheadGemini 3.8 Live is clearly ahead in 0 of 6 subtests and slightly in 4, Gemini 3.8 Live Extended Thinking in 1 clearly and slightly in 1.
Artificial Analysis Big Bench Audio speech reasoning
1,000 tasks, as of October 15, 2026
Artificial Analysis τ-Voice task completion
as of October 15, 2026
Artificial Analysis Time to First Audio
lower is better, as of October 17, 2026
Show all 6 subtests
| Subtest | Gemini 3.8 Live | Gemini 3.8 Live Extended Thinking | Verdict |
|---|---|---|---|
| Artificial Analysis Big Bench Audio speech reasoning | 92.0 % | 98.0 % | Gemini 3.8 Live Extended Thinking ahead |
| Artificial Analysis τ-Voice task completion | 30.1 % | 68.6 % | Gemini 3.8 Live Extended Thinking slightly ahead |
| Artificial Analysis Time to First Audio (lower is better) | 1.18 s | 1.35 s | Gemini 3.8 Live slightly ahead |
| Artificial Analysis Speech Agent Arena preference Elo | 1,079 | 970 | Gemini 3.8 Live slightly ahead |
| Speech Agent Arena task success rate | 93.2 % | 89.1 % | Gemini 3.8 Live slightly ahead |
| Artificial Analysis Conversational Dynamics | 96.1 % | 91.9 % | Gemini 3.8 Live slightly ahead |
Source · latest October 28, 2026
Cekura
Gemini 3.8 Live Extended Thinking slightly aheadGemini 3.8 Live Extended Thinking is clearly ahead in 0 of 4 subtests and slightly in 3, Gemini 3.8 Live in none. 1 is level.
Cekura TurnBench end-of-turn score, Light noise
36 tasks, as of October 23, 2026
Cekura TurnBench end-of-turn score, Loud noise
36 tasks, as of October 23, 2026
Cekura TurnBench end-of-turn score, Echoey rooms
36 tasks, as of October 23, 2026
Show all 4 subtests
| Subtest | Gemini 3.8 Live | Gemini 3.8 Live Extended Thinking | Verdict |
|---|---|---|---|
| Cekura TurnBench end-of-turn score, Light noise | 77.0 % | 81.0 % | Gemini 3.8 Live Extended Thinking slightly ahead |
| Cekura TurnBench end-of-turn score, Loud noise | 78.0 % | 82.0 % | Gemini 3.8 Live Extended Thinking slightly ahead |
| Cekura TurnBench end-of-turn score, Echoey rooms | 83.0 % | 86.0 % | Gemini 3.8 Live Extended Thinking slightly ahead |
| Cekura TurnBench end-of-turn score (six-condition average) | 82.0 % | 84.0 % | level |
Source · latest October 23, 2026
All data side by side
Prices are verbatim from the provider, because audio models bill per minute, character or token.
| Gemini 3.8 Live | Gemini 3.8 Live Extended Thinking | |
|---|---|---|
| Task | Real-time voice dialog | Real-time voice dialog |
| Developer | ||
| Status | Active | Active |
| Weights | proprietary | proprietary |
| Runs locally | no | no |
| Price | $0.75 per 1M text tokens | $0.75 per 1M text tokens |
| Popularity (out of 100) | 0 | 0 |