Memperkenalkan Gemini 3.8 Live dan 3.8 Live Extended Thinking: Google Membagi Lini Suaranya Menjadi Dua

Google's Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking arrived on September 15, 2026 — one announcement, two models, and a genuine fork in the product line. The headline number is a 6.6-point gap on Artificial Analysis's Speech to Speech Index, 82.6 to 76.0. The number underneath it is 38. On the same board's τ-Voice agentic task completion measure, the reasoning variant resolves 68.6% of replica customer-service scenarios and the standard model resolves 30.1%. The 6.6 points are what the index says. The 38 points are what happens when you ask either model to actually get something done.

A day on from launch, the fork has a price attached. Both models carry published per-minute audio rates through the Live API, and Artificial Analysis's leaderboard puts cost-per-hour of input audio at $0.84 for the standard model against $3.50 for the reasoning variant — a little over four times as much per hour for those 6.6 index points. That trade is the whole decision, and reading it off the index alone gets it wrong.

Yang membuat rilis ini berbeda dari penyegaran model suara biasa adalah bahwa pembagiannya bukanlah tingkatan ukuran. Ini adalah ketidaksepakatan tentang apa tugas sebuah agen suara. Salah satu dari model ini adalah antarmuka percakapan. Yang lainnya adalah sistem penalaran yang kebetulan berbicara.

Dua model, dua pekerjaan

Kerangka Google sendiri luar biasa jelas tentang hal ini. Gemini 3.8 Live dijelaskan sebagai "dibangun untuk skala dan efisiensi biaya, menggabungkan kecerdasan percakapan dengan dialog yang lancar dan landasan visual." Gemini 3.8 Live Extended Thinking adalah "dibangun untuk tugas-tugas kompleksitas tinggi, dengan kecerdasan yang meningkat dan penalaran multi-langkah."

Perbedaan praktisnya terlihat pada daftar kemampuan, bukan pada lembar spesifikasi:

• Gemini 3.8 Live — pemrosesan input visual yang hampir real-time, pengalihan otomatis di tengah percakapan di antara 97 bahasa yang didukung, serta eksekusi tool/API di latar belakang sehingga model mengakui permintaan dan terus berbicara sementara panggilan tersebut diselesaikan.

• Gemini 3.8 Live Extended Thinking — bernalar dan berbicara pada saat yang sama, menuturkan kemajuannya dengan petunjuk seperti "Coba saya periksa…" selagi tugas bertahap berjalan.

• Dibagikan — semua audio yang dihasilkan membawa watermark SynthID milik Google, keduanya tercakup dalam kartu model yang diterbitkan dengan nama gemini-3-8-audio, dan keduanya kini memiliki ID model API yang terdokumentasi: gemini-3.8-live dan gemini-3.8-live-extended-thinking.

Model Extended Thinking bukan sekadar 3.8 Live dengan anggaran berpikir yang lebih panjang. Model ini adalah model yang dapat mempertahankan percakapan dan rencana tugas secara bersamaan tanpa percakapan menjadi macet — yang justru merupakan hal yang membuat agen suara gagal di produksi. Pengguna tidak ingin ada keheningan saat agen mencari sesuatu. Mereka ingin agen terus berbicara.

The board says 6.6. The components say 38

The Speech to Speech Index is a weighted average of four underlying results: Speech Reasoning, measured by Artificial Analysis's own Big Bench Audio set; Agentic Performance, measured by running the τ-Voice customer-service benchmark; Arena Preference, taken from the Speech Agent Arena; and Task Success Rate. Only models with all four components receive an index score, which is why some rows on that board are blank.

Pulled apart, the two new models do not look like a 6.6-point pair at all:

• Speech to Speech Index — Gemini 3.8 Live Extended Thinking (High) 82.6 vs Gemini 3.8 Live 76.0

• Agentic performance (τ-Voice task completion) — 68.6% vs 30.1%

• Speech reasoning (Big Bench Audio) — 98% vs 92%

• Conversational dynamics — 91.9% vs 96.1%

• Arena preference (Elo) — 990 vs 1083

• Task success rate — 89.1% vs 93.2%

• Time to first audio — 1.35s vs 1.18s

• Cost per hour of input audio — $3.50 vs $0.84

Read it honestly and the standard model wins four of eight lines. It is faster to first audio, rated higher by the arena, more likely to complete a task, and rated better on conversational dynamics. It is also a third of the price. What it cannot do is finish the job when the job has steps: 30.1% against 68.6% is the largest single gap anywhere in this release, and it sits on the one axis that separates a voice agent from a voice interface.

Two caveats before that table gets treated as settled. The arena preference figure used in the index is frozen at the point a model becomes eligible for publication, so it does not track the live Elo on the Speech Agent Arena chart — read it as a snapshot, not a running score. And the τ-Voice margins at the top of the board are thin enough to be noise: the reasoning variant's 68.6% leads GPT-Live-1 at 67.9% (Astra backend, medium effort) by 0.7 points, with GPT-Live-1 (Sol, low) at 59.3% and Grok Voice Think Fast 2.0 High at 56.5% behind it. The 38-point gap between the two Gemini models is not thin. The 0.7-point lead over GPT-Live-1 is.

Where it sits on the index

In the capture below, taken on September 16, 2026, the top of the board reads like this:

• Gemini 3.8 Live Extended Thinking (High) — 82.6 (peringkat pertama)

• GPT-Live-1 (backend Astra, upaya sedang) — 81.5

• Grok Voice Think Fast 2.0 High — 81.3

• GPT-Live-1 (backend Sol, upaya rendah) — 80.1

• Gemini 3.8 Live — 76.0

• GPT-Realtime-2.1 Tinggi — 73.9

• Gemini 3.1 Flash Live High — 71.5

That is the case for the split in one screen. The Extended Thinking variant takes the top spot, running at the board's (High) reasoning-effort label — the same convention it uses for Grok Voice Think Fast 2.0 High and GPT-Realtime-2.1 High, and the reason the "(High)" suffix you may see attached to this model in coverage is a configuration label rather than a separate release. The standard variant lands fifth — below two GPT-Live-1 configurations and below Grok. Google's own blog describes the standard model as "highly cost-effective" and notes it "secured a second place in the Speech Agent Arena," which is a different board measuring a different thing. Both statements are true. Read together they say: 3.8 Live is a very good conversational model at a very good price, and it is not the frontier of voice intelligence.

Kebocoran yang tiba lebih dulu

Model-model itu terlihat di halaman kuota Google Cloud sehari sebelum pengumuman, dan kami meliput penampakan tersebut saat itu sebagai slug yang belum dikonfirmasi tanpa kartu model, tanpa harga, dan tanpa konfirmasi Google. Postingan itu benar soal statusnya dan salah soal linimasa — konfirmasinya datang dalam waktu sekitar 24 jam.

Pelajaran ini layak dicatat, karena bertentangan dengan naluri yang biasa. Slug yang bocor sehari sebelum peluncuran adalah peluncuran. Slug yang bocor tanpa dokumen pendukung selama delapan minggu adalah hal yang sama sekali berbeda. Sinyal yang membedakan keduanya bukanlah slug; melainkan keberadaan halaman kuota, yang hanya ada ketika kapasitas telah disediakan.

Berapa biayanya, per menit

Saat pengumuman, materi Google tidak memisahkan kedua model dari segi harga — model standar diberi tarif sementara varian penalaran tidak, dan itulah yang dicatat oleh papan skor di atas. Kesenjangan itu kini sudah tertutup. Kedua model kini memiliki harga yang dipublikasikan melalui Live API: $0,005 per menit untuk input audio dan $0,018 per menit untuk output audio, yang dikonfirmasi Google pada 16 September saat keduanya diluncurkan. Kartu tarif terbitan yang sama juga mencakup harga token untuk tier Gemini 3 Live — $0,75 per 1 juta token teks masuk dan $4,50 keluar, $3,00 per 1 juta token audio masuk dan $12,00 keluar, serta $1,00 per 1 juta untuk input gambar atau video — meskipun baris-baris token tersebut dipublikasikan untuk tier, bukan per model, jadi bacalah sebagai kartu tarif tier tersebut dan bukan sebagai penawaran harga untuk varian tertentu.

Artificial Analysis has also filled in the number that matters most for planning. Its leaderboard now carries a cost-per-hour-of-input-audio column — the cost to complete a fixed 40-question Big Bench Audio subset, normalised to an hourly rate — and both new models have a figure there:

• Gemini 3.8 Live — $0.84/hour of input audio, the lowest paid rate on that board

• Gemini 3.8 Live Extended Thinking (High) — $3,50/jam, masih di bawah kedua model yang dikalahkannya dalam hal kualitas

• Grok Voice Think Fast 2.0 High — $4,80/jam

• GPT-Live-1 (backend Astra, upaya sedang) — $5,83/jam

• GPT-Realtime-2 (Tinggi) — $4,14/jam

• GPT-Realtime-2.1 High — $10.75/jam

One note on the captures above, since the costs panel in the leaderboard screenshot predates these two rows: neither $0.84 nor $3.50 appears in it, and its cheapest bar is $1.42. The figures in the list are read from the board's summary table as it stands on September 16, 2026, not from that image. The index values shown in the image are unaffected and match the table.

Read the two new numbers against the components above and the release stops being a two-model announcement and becomes a single decision with a published exchange rate. The reasoning variant buys 6.6 index points, six points of speech reasoning and 38.5 points of agentic task completion for a little over four times the hourly audio cost. Whether that is worth paying depends entirely on what your agent is doing — and the shape of the answer changed once the components were published. A voice agent whose job is to finish a multi-step task is buying the single largest improvement in this release, at $3.50 an hour while undercutting GPT-Live-1 Astra by about 40% and Grok Voice Think Fast 2.0 High by roughly a quarter, and scoring above both. A voice agent whose job is to converse — answer, hold a thread, take a message, be pleasant — is buying very little with extra reasoning depth, and at $0.84 an hour it is buying the cheapest competent voice model anyone currently publishes.

Itulah bentuk daftarnya. Google telah menetapkan harga model penalarannya di bawah dua model yang dikalahkannya, dan menetapkan harga model standarnya di bawah semua yang ada di papan. Jika Anda menjalankan menit suara dalam volume besar, angka inilah yang menggerakkan tagihan Anda, bukan skor indeks — dan fakta bahwa kedua varian itu terpaut 4× berarti memilih mana yang akan dipanggil adalah pengungkit biaya tunggal terbesar di tumpukan.

"Private preview" benar-benar berperan penting dalam kalimat itu

Tidak satu pun dari model tersebut tersedia secara umum dalam pengertian kontraktual, tetapi permukaan pengembangnya terbuka dan kini memiliki harga, yang mengubah apa yang dapat Anda rencanakan:

• Gemini 3.8 Live — pengembang mendapatkannya di Gemini API, Live API, dan Google AI Studio dengan ID terdokumentasi gemini-3.8-live; perusahaan mendapat pratinjau privat di Gemini Enterprise, dengan Gemini Enterprise for Customer Experience "segera hadir"; semua orang mendapatkannya di Search Live.

• Gemini 3.8 Live Extended Thinking — permukaan pengembang yang sama di bawah gemini-3.8-live-extended-thinking, plus jalur konsumen yang lebih luas: Gemini Live, Docs Live untuk pelanggan Google AI Pro dan Ultra, serta Gmail Live dan Keep Live untuk semua pelanggan Google AI.

Yang masih hilang adalah bagian yang dibutuhkan perusahaan besar: komitmen ketersediaan umum, angka uptime yang dipublikasikan, dan tarif yang dijanjikan Google untuk dipertahankan. Jika Anda membutuhkannya, tunggu. Jika Anda sedang membuat prototipe, jalur ini terbuka hari ini dengan daftar harga nyata yang mendukungnya, yang merupakan posisi secara material lebih baik daripada pratinjau tanpa harga yang dipublikasikan. Membangun integrasi sekarang masuk akal; menempatkan saluran telepon yang kritis terhadap pendapatan pada target ketersediaan yang belum dijanjikan Google tidaklah masuk akal, dan solusi untuk itu dijelaskan di bawah.

Di mana klaim vendor berakhir

Google published per-component figures of its own alongside the launch — 68.6% on τ-Voice, 35.1% on Sierra's τ³-Banking leaderboard, and 97.7% on Big Bench Audio, all credited to the Extended Thinking model. Those now divide into two groups, and the split matters.

Two of the three line up with measurements Artificial Analysis runs itself under its own harness. Its τ-Voice agentic component reads 68.6% for this model, against 67.9% for GPT-Live-1 Astra and 56.5% for Grok Voice Think Fast 2.0; its Big Bench Audio speech-reasoning component reads 98%, against Google's 97.7%. Artificial Analysis describes τ-Voice and Big Bench Audio as its own benchmarks, run across three trials where available. So the τ-Voice and reasoning numbers now sit on a public board you can go and read, not only in Google's announcement — treat them as corroborated rather than settled, since the exact figures Google quoted may well be that same run rather than a second one.

The Sierra number has no such backing and remains a vendor claim: 35.1% on Sierra's τ³-Banking leaderboard against 32.0% for GPT-Live-1 Astra and 16.5% for xAI-Realtime. It is also the number worth sitting with, because 35.1% means the leading voice model on this board fails roughly two of every three realistic banking task-completion attempts. Google also cites a ServiceNow EVA-Bench run performed on the Live API in Gemini Enterprise Agent Platform, and describes both models as pushing the Pareto frontier for complex workflows — neither of which has an independent reading attached.

There is a precedent worth holding onto here. xAI reported a Speech to Speech Index figure of 82.9 for Grok Voice Think Fast 2.0 in July. On the board captured above, that model sits at 81.3. Vendor-reported index scores do not always survive contact with a live leaderboard — usually because the index is revised, sometimes because the configuration tested was not the one that shipped. Verify the τ-Voice and Big Bench numbers on your own traffic, and read the agentic column as the one to test hardest.

Lapisan tempat agen-agen ini sebenarnya berjalan

Kedua model berada di depan sebuah backend. Eksekusi alat di latar belakang, perencanaan tugas multi-langkah, dan pola delegasi yang digunakan setiap agen suara serius semuanya bermuara pada panggilan model teks biasa — dan itulah lapisan dengan variasi biaya paling besar dan keterikatan paling sedikit.

OrcaRouter menyediakan 190 model dari satu kunci dengan harga daftar penyedia tanpa markup, yang berarti pemotongan harga dari lab hulu berlaku di sisi kami pada hari yang sama, bukan pada perpanjangan kontrak berikutnya. Untuk stack suara yang front-end-nya adalah model pratinjau, dua properti yang berguna adalah target delegasi dapat ditukar tanpa menyentuh integrasi suara, dan failover otomatis menjaga panggilan tetap hidup saat backend mengalami error atau timeout. Untuk lebih tepat tentang apa yang kami host dan tidak kami host: endpoint Gemini 3.8 Live tidak ada di router kami — itu berasal dari API Google sendiri — tetapi model teks yang sangat sering menjadi tempat agen-agen ini mengalihkan pekerjaan adalah, Gemini 3.8 Flash termasuk di antaranya.

Tontonan Berikutnya

Three things will settle the open questions in this release. The first is general availability and a rate Google has committed to hold — the prices are published now, but preview pricing has a habit of moving, and a rate card is not a contract. The second is whether the arena and task-success columns hold up, because the standard model's 1083 Elo and 93.2% task success are the strongest argument against paying 4× for the reasoning variant, and the index's arena figure is frozen at eligibility rather than live. The third is independent runs on the agentic gap itself: 68.6% against 30.1% is the largest claim in the release and the one most worth reproducing, because everything else about the two models is close.

Until then, the honest summary runs on two numbers rather than one. Gemini 3.8 Live Extended Thinking is the best-scoring voice model on the public board, leads the agentic component outright, and undercuts the two models directly behind it on hourly cost. Gemini 3.8 Live is the cheapest competent voice model on that board, is preferred by the arena and more reliable on shallow tasks, and sits 6.6 points back overall with a hard ceiling at the one thing agents get hired for. Google has shipped the same fork twice, four times apart on price, and made the choice unusually easy to price — provided you read the components and not just the index.