Gemini 3.8 Live
Sep 2026
Gemini 3.8 LiveCompare inon
Gemini 3.8 Live measured on 15 of 34 Realtime benchmarks15
BenchmarkByScoreRank
Agents4
Multimodal9
Legacy · saturated or no longer reported2
No benchmark matches
$0.005 / $0.018per minute in / out
1.2 sfirst audio?
30.1τ-Voice
Sep 2026released
en, es, fr +97Languages
Provider-stated input/output audio minute components; text, tools and session/backend charges remain separate.Price covers
Ranking?
Full ranking →···
12GPT Realtime 2.195.813Gemini 3.8 Live91.714GPT Live 190.1Strong at
Alternatives?
ModelScoreCostPriceWhat changes
BetterEleven Realtime v1 Max+8%—$0.08Languages: not verified
OpenStep-Audio R1.1+6%——Languages: not verified
BetterGrok Voice 3.0+7%—$0.08Languages: loses es, zh-Hans, zh-Hant +83BetterGrok Voice 2.0unstated+6%—$0.08Languages: not verifiedBetterStepAudio 3 Realtime Preview+9%——Languages: not verifiedModelScoreCostPriceWhat changes
Nothing here yet.
ModelScoreCostPriceWhat changes
Eleven Realtime v1 Max+8%—$0.08Languages: not verifiedGrok Voice 3.0+7%—$0.08Languages: loses es, zh-Hans, zh-Hant +83Grok Voice 2.0unstated+6%—$0.08Languages: not verifiedStepAudio 3 Realtime Preview+9%——Languages: not verified
Qwen Audio 3.0 Realtime Plus+8%——Languages: not verifiedModelScoreCostPriceWhat changes
Step-Audio R1.1Open+6%——Languages: not verifiedKRaon SpeechChat 9BOpen-37%——Languages: not verifiedQwen3 Omni 30B A3B InstructunstatedOpen-38%——Languages: loses es, fr, de +96
NemotronLabs VoiceChat 11BOpen-71%——Languages: loses es, fr, de +96PersonaPlex 7BOpen-79%——Languages: loses es, fr, de +96Benchmarks
in Speech & VoiceBig Bench Audio · AA v1.291.7#13
Full-Duplex-Bench · AA subset v1/v1.596.1#6
Time to first audio · AA1.2#17
AA Speech Agent Arena · overall · Elo1,077.6#2
AA Speech Agent Arena · overall · Task Success93.2#3
AA Speech Agent Arena · agentic · Elo1,064.1#3
AA Speech Agent Arena · non-agentic · Elo1,119.7#2
τ-Voice · AA · aggregate30.1#12
τ-Voice · AA · airline48.7#10
τ-Voice · AA · retail29.2#11
τ-Voice · AA · telecom12.3#14
EVA-Bench · EVA-A accuracy pass45.9#7
EVA-Bench · EVA-X experience pass88.6#1
EVA-Bench · response speed2#5
FAQ
Is Gemini 3.8 Live free to use?›
There is no free API; the cheapest paid price is $0.005 per 1M input tokens.
How good is Gemini 3.8 Live, and what is it best at?›
#12 of 17 in Realtime.
Is Gemini 3.8 Live better than Gemini 2.5 Flash Exp Native Audio Thinking Dialog?›
On Big Bench Audio · AA v1.2, Gemini 3.8 Live scores 91.7 vs 90.7 — higher.
How can I use Gemini 3.8 Live?›
Through the Google API.
When was Gemini 3.8 Live released? Is there a newer version?›
Released 2026-09-15. It is the current version in its line.
Is Gemini 3.8 Live open source?›
No, Gemini 3.8 Live is a closed model available through an API. The closest open-weights option here is Step-Audio R1.1.
What is Gemini 3.8 Live?›
Gemini 3.8 Live is a closed speech & voice model from Google, released Sep 2026, the successor to Gemini 2.5 Flash Exp Native Audio Thinking Dialog.
What are the best alternatives to Gemini 3.8 Live?›
Eleven Realtime v1 Max (better), Step-Audio R1.1 (open weights).