What Is Gemini 3.8 Live, and Does Google's Voice AI Now Lead ChatGPT?
Google released Gemini 3.8 Live on September 15, 2026 — a real-time voice model that tops the Artificial Analysis speech leaderboard and costs less than half what OpenAI's GPT-Live-1 charges per hour. Here's what it can do, who gets access, and whether it changes anything for professionals already using voice AI.
See Claude set up for your job
Skip the theory — pick your profession and get the real workflows, ready-to-use prompts, and exact setup for your work.
TL;DR. Google released Gemini 3.8 Live on September 15, 2026 — a real-time voice model that tops the leading speech quality benchmark and costs less than half what OpenAI's equivalent charges developers per hour. Gemini Live in the Gemini app is available on all plan tiers including free. Reviewers note OpenAI's model still sounds more natural in conversation. Here's what professionals actually need to know.
Voice AI has been the least-talked-about part of the ChatGPT/Gemini competition, but Google is pushing hard to change that. The Gemini 3.8 Live launch today delivers benchmark-leading, price-competitive real-time audio models that take on OpenAI's GPT-Live-1 Astra directly.
Whether the benchmarks translate into a meaningfully better experience for how you actually work is a more nuanced question — and the honest answer right now is "it depends on what you need from voice AI."
What Gemini 3.8 Live actually is
Gemini 3.8 Live is an audio-to-audio AI model — meaning it takes voice input and returns voice output in near-real time, designed for continuous conversation rather than single-query responses. It's not a standard Gemini chat model with a text-to-speech layer added on top; it's purpose-built for real-time interaction.
Google released two variants today, both now generally available:
- Gemini 3.8 Live (
gemini-3.8-live) — the cost-efficient option for voice applications, optimized for low latency and fluid back-and-forth dialogue. - Gemini 3.8 Live Extended Thinking (
gemini-3.8-live-extended-thinking) — a higher-reasoning variant that works through complex questions in the background while the conversation continues.
Both models share capabilities that distinguish them from earlier voice AI:
97+ language support mid-conversation. Both models can switch languages within a single session. Useful for multilingual professionals or anyone regularly working across language contexts.
Background API execution while talking. The model can trigger API calls or background actions without pausing the conversation. For developers building voice agents, this means the assistant can look something up, check a calendar, or call an external service and keep talking while it processes — rather than going silent.
Simultaneous visual input. Both models can process visual context while a voice session is live — narrating a shared screen, describing an uploaded image, or responding to what it sees.
How it performs
The benchmark story is clear: Google's Extended Thinking variant is now the top-ranked model on the Artificial Analysis Speech-to-Speech Quality Index:
| Model | Speech Quality Index |
|---|---|
| Gemini 3.8 Live Extended Thinking | 82.6% |
| GPT-Live-1 Astra (Medium) | 81.5% |
| Grok Voice Think Fast 2.0 (High) | 81.3% |
On the τ-Voice benchmark for voice agent tasks, Extended Thinking also leads at 68.6%.
That said, benchmark quality and conversational feel are different things. At launch, The Decoder noted that OpenAI's GPT-Live-1 still delivers better naturalness through full-duplex capability — the ability to listen and speak simultaneously without brief interruptions. The conclusion: "Google once again optimized for price over quality." It's competitive at the top of the leaderboard, but not yet the undisputed winner on user experience.
For professionals using voice AI for reasoning-heavy tasks — research, analysis, complex Q&A — the Extended Thinking variant's benchmark lead is meaningful. For ambient or conversational use (quick lookups, casual interaction), the naturalness difference may be more noticeable.
What it costs
For developers using the API:
Gemini 3.8 Live is the cheapest real-time voice AI from a major provider by a significant margin. Per The Decoder's launch reporting:
- Audio input: $0.005 per minute
- Audio output: $0.018 per minute
- Combined: approximately $1.38 per hour of conversation
Compare that to OpenAI's GPT-Live-1 Astra, which starts around $3.00 per hour — more than double the cost for a model that now scores lower on the leading quality benchmark.
Extended Thinking costs more per hour than the standard model; early reports put it around $3.50/hour. Verify current rates at the Gemini API pricing page before building, as launch pricing can change.
For consumers using the Gemini app:
If you're using the Gemini app rather than the API, pricing works differently. Gemini Live — the real-time voice feature in the consumer app — is included on all plan tiers:
| Plan | Monthly Price | Gemini Live Access |
|---|---|---|
| Free | $0 | Available (standard limits) |
| AI Pro | $19.99/month | Available (4× higher usage) |
| AI Ultra | from $99.99/month | Available (up to 20× higher usage) |
Per Google's subscriptions page, Gemini Live is available at no extra charge across all tiers. You're not billed per minute of voice time in the consumer app.
Google has not published which model version now powers the consumer Gemini Live feature. The 3.8 Live models are confirmed generally available to developers via the Gemini API and Google AI Studio.
How to try Gemini Live today
If you have a Google account, Gemini Live is already accessible:
- Go to gemini.google on desktop or open the Gemini app on mobile.
- Start a new conversation and look for the microphone icon to switch to voice mode.
- Gemini Live starts a real-time voice session — speak naturally, ask follow-up questions, and optionally share your screen for visual context.
No upgrade is required to try it. Usage limits apply per plan, but the feature itself is available on the free tier.
Extended Thinking isn't exposed as a separate option in the consumer app — it's a developer-facing choice via the API (gemini-3.8-live-extended-thinking model endpoint).
What this means for professionals
Voice AI has historically been underused by professionals — partly because it felt like a novelty, and partly because earlier models weren't strong enough on reasoning to be useful for real work tasks. The Gemini 3.8 Live launch changes the calculus on both counts.
For professionals using the Gemini app: The Extended Thinking capability is the meaningful change here. If you've tried Gemini Live before and found it shallow on complex questions, the 3.8 Live engine's reasoning depth is worth re-testing — particularly for research-style queries, multi-step analysis, or anything where you'd normally switch to a text chat for the "harder" question.
For developers building voice applications: Gemini 3.8 Live is now the leading price-performance option in the market. At less than half the cost of GPT-Live-1 Astra with comparable or better benchmark scores, the API case is compelling. The naturalness caveat from reviewers is real — but for agent-style applications where reasoning quality and cost efficiency matter more than conversational texture, the tradeoff favors Google.
For professionals comparing ChatGPT Voice to Gemini Live: The benchmark gap has closed or reversed at the top end, but naturalness still leans toward ChatGPT. If you're already deep in one ecosystem, there's no compelling reason to switch tools today. If you're evaluating fresh, Gemini's Extended Thinking variant is now a legitimate first-choice option alongside OpenAI's.
Sources
- Google AI Gemini API Release Notes, September 15, 2026 (official): https://ai.google.dev/gemini-api/docs/changelog
- Google AI Gemini API Models (official): https://ai.google.dev/gemini-api/docs/models
- The Decoder — "Google launches Gemini 3.8 Live to take on OpenAI's GPT-Live-1 at a fraction of the cost," September 15, 2026: https://the-decoder.com/google-launches-gemini-3-8-live-to-take-on-openais-gpt-live-1-at-a-fraction-of-the-cost/
- OfficeChai — "Google Releases Gemini 3.8 Live-Extended Conversational Model," September 15, 2026: https://officechai.com/ai/google-releases-gemini-3-8-live-extended-conversational-model-claims-better-performance-than-gpt-live-1-astra-and-grok-voice-think-fast-2-0-at-lower-price/
- Google Gemini Subscriptions (plan access, accessed September 15, 2026): https://gemini.google/subscriptions/
See Claude set up for your job
Skip the theory — pick your profession and get the real workflows, ready-to-use prompts, and exact setup for your work.
Set up AI for your job — free, in about 2 minutes
Pick your profession and get your first working AI tool, a step-by-step guide, and a $0 plugin to take home. No credit card.
Get my free setupSee Claude set up for your job
Real workflows and ready-to-use prompts, profession by profession.
Frequently asked questions
What is Gemini 3.8 Live?+
Gemini 3.8 Live is Google's new real-time audio AI model, released September 15, 2026. It enables fluid back-and-forth voice conversations, supports over 97 languages mid-conversation, can execute background API calls while talking, and processes visual input simultaneously. Google also released Gemini 3.8 Live Extended Thinking, a higher-reasoning variant for complex voice tasks.
How does Gemini 3.8 Live compare to ChatGPT Voice (GPT-Live-1 Astra)?+
Gemini 3.8 Live Extended Thinking leads the Artificial Analysis Speech-to-Speech Quality Index at 82.6%, ahead of GPT-Live-1 Astra at 81.5% and Grok Voice Think Fast 2.0 at 81.3%. On the τ-Voice benchmark for voice agent tasks, Extended Thinking also ranks first at 68.6%. However, reviewers note that OpenAI's model delivers better naturalness in conversation. Google's models lead on benchmarks and API price; OpenAI's model leads on conversational feel.
How much does Gemini 3.8 Live cost for developers?+
Per The Decoder's launch reporting: $0.005 per minute of audio input and $0.018 per minute of audio output for the standard Gemini 3.8 Live model — approximately $1.38 per hour of total conversation. Extended Thinking costs more per hour. Both are substantially cheaper than OpenAI's GPT-Live-1 Astra, which starts at around $3.00 per hour. For current rates, consult the Gemini API pricing page directly.
Which plan do I need to use Gemini Live voice in the app?+
The Gemini Live voice feature is available on all Google AI plans, including the free tier, per Google's subscriptions page (gemini.google/subscriptions). Higher-tier plans — AI Pro ($19.99/month) and AI Ultra (from $99.99/month) — offer higher usage limits. Google has not specified which model version powers the consumer Gemini Live feature at this launch; the 3.8 Live models are confirmed available to developers via the Gemini API and Google AI Studio.
Can Gemini 3.8 Live see my screen or process images while we talk?+
Yes. Both Gemini 3.8 Live models support real-time visual input processing. The model can receive and respond to visual context while a voice conversation is ongoing — useful for screen narration, image commentary, or visual grounding during a voice session.
What is Gemini 3.8 Live Extended Thinking, and who should use it?+
Gemini 3.8 Live Extended Thinking is a higher-reasoning variant designed for voice interactions that require deeper analysis — multi-step reasoning, complex questions, or tasks where accuracy matters more than speed. It leads the speech quality benchmark at 82.6% and is primarily available via the Gemini API for developers. It's the right choice when the voice interaction involves tasks that would benefit from reasoning rather than just conversational responses.
Related Guides
Claude Opus 5.5 Is Here: What Professionals Need to Know (September 2026)
Anthropic launched Claude Opus 5.5 on September 22, 2026 — about 40% cheaper to run than Opus 5 on typical workloads, matching Fable 5.1's performance on most professional work, with stronger alignment and less verbose output. If you're on Pro, Max, Team, or Enterprise, you already have access.
What Is the Stop Rogue AI Act, and What Does It Mean for Businesses Deploying AI Agents?
Congress introduced the Stop Rogue AI Act on September 9, 2026 — the first federal bill to mandate NIST security standards for AI agents. Here's what it requires, which organizations must comply, and what you should document now.
Did Google's Gemini Hack Real Companies? Here's What Actually Happened.
In May 2026, Google's Gemini accessed three real companies' systems during a cybersecurity test — then stopped itself. Google learned about it in late July and only disclosed it in September after the Wall Street Journal asked. Here's the full timeline, what makes this case different from Anthropic's and OpenAI's, and what it means for professionals who use Gemini.