Gemini in October 2026: 3.8 Live, 3.8 TTS, Omni in Vids, Gemini 4 Argon, and Nano Banana 2.1

September 3 through October 8

Since our September update, Google shipped several GA models, closed one Omni preview endpoint, and announced a new flagship text model that is not yet available to developers. This post covers September 3 through October 8, 2026, with emphasis on changes that still affect production systems and hard deadlines you should track.

Omni updates

Omni 1.1 Flash in Google Vids (September 23)

Per Google Workspace Updates, Omni 1.1 Flash entered Google Vids on September 23. Compared with earlier Vids video generation, Google highlights:

Capability	Notes
Video extension	Higher quality; improved character and audio consistency
Duration	Specify exact clip length
Resolution	Generate 1080p directly, or upscale AI video to 1080p

Availability: Business Starter, Standard, and Plus; Enterprise Starter, Standard, and Plus; Education Plus; and personal Google accounts with Google AI Pro or Ultra. There is no admin toggle—the feature rolls out to eligible accounts directly.

Preview endpoint shutdown (September 30)

gemini-omni-flash-preview shut down on schedule on September 30, 2026. Use the GA model gemini-omni-1.1-flash for all Omni Flash API and production workloads.

Omni Pro

As of October 8, Google has not published a GA date for Gemini Omni Pro. The Google AI subscription plans page currently lists Omni Flash for the Gemini app and Flow; it does not list Omni Pro. We have seen no official update beyond the existing Ultra-first preview narrative from earlier months.

Gemini 3.8 Live (September 15 GA)

Google GA’d the Gemini 3.8 Live family on September 15, 2026 for low-latency voice agents and live audio interaction.

Model ID	Role
gemini-3.8-live	Default for low-latency voice agents; interleaved reasoning; async function calling on by default; full-session client content updates
gemini-3.8-live-extended-thinking	Higher reasoning; background reasoning during live audio interactions, recommended for complex tasks

Developers reach both models through the Gemini API and Google AI Studio. Enterprises can request Gemini Enterprise private preview. Consumers use Search Live. Extended Thinking is also available in Gemini Live and in Docs, Gmail, and Keep.

gemini-3.8-live is now the default choice for Live API voice agents, ahead of gemini-3.1-flash-live-preview. If you still target the 3.1 preview, plan a migration and review integration patterns in our Gemini 3.1 Flash Live API guide.

Gemini 3.8 TTS (September 22 API GA)

The Gemini 3.8 Flash TTS line reached API GA on September 22; Google’s official blog post followed on September 23.

Model ID	Use case
gemini-3.8-flash-tts	Creative direction, character design, line-by-line performance control, dialects, stable multi-turn long-form speech
gemini-3.8-flash-lite-tts	High throughput, lower cost; intended replacement for gemini-3.1-flash-tts-preview

New Voices API surface: /v1beta/voices supports voice design, voice cloning (with consent verification), and 150+ preset and custom voices. Gemini 3.8 Flash-Lite TTS is also available in Google Vids; Gemini 3.8 Flash TTS is available in the Gemini API and Google AI Studio, with Gemini Enterprise API access listed as coming soon.

Lyria 3.5 (September 3 GA)

lyria-3.5 GA’d on September 3 for full-song generation. Google cites better musical coherence, more natural vocals, and finer control over duration and structure. Inputs: text and image. Output: 44.1 kHz stereo audio.

Gemini 4 Argon (announced September 30)

On September 30, Google announced Gemini 4 Argon, a new flagship model. Developers cannot use it today. Access is limited to trusted cybersecurity defenders through the Fairwind program while Google participates in the U.S. government’s voluntary pre-release model access process.

After that process, Google plans to open Argon first to paid API customers and Google AI Ultra subscribers. No public availability date has been announced.

Introductory pricing (while the intro period lasts):

	Input (per 1M tokens)	Output (per 1M tokens)
Intro period	$2	$10
Cached input	95% discount on cached input tokens	—
After intro period	$4	$20

Do not slot Argon into production roadmaps until Google publishes model IDs and general API access. For earlier context on the Gemini 4 line, see our Gemini 4 roadmap preview.

Nano Banana 2.1 (October 6 GA)

gemini-nano-banana-2.1 reached GA on October 6 for image generation and conversational editing. Google reports better image quality, prompt adherence, multi-turn character consistency, and text rendering.

New aspect ratios: 1:4, 4:1, 1:8, 8:1, with 1K, 2K, and 4K output sizes.

gemini-3.1-flash-image is deprecated and shuts down on October 29, 2026. Migrate before that date.

from google import genai

client = genai.Client()

response = client.models.generate_content(
    model="gemini-nano-banana-2.1",  # was gemini-3.1-flash-image
    contents="A product photo on a white background, soft shadow.",
)

Other API changes

Date	Change
September 1	Agentic video understanding on gemini-3.7-flash, gemini-3.6-flash, and gemini-3.5-flash-lite: long-video understanding with up to 88% fewer tokens
September 17	antigravity-preview-09-2026 replaces antigravity-preview-05-2026; the May preview shut down October 5
Antigravity local tools	Parameter names moved from snake_case to PascalCase; file edits use line-range replacements instead of full-file rewrites
Antigravity remote sandbox	If you only read output_text / model_output, update the agent model string; no local tool parser changes required

Gemini 2.5 status correction

Google updated its guidance on September 18: the Gemini 2.5 family is not deprecated. Models remain available until further notice, but access is limited to accounts that already used 2.5. New projects should use gemini-3.5-flash-lite or gemini-3.8-flash instead.

Our September 2026 update cited an October 16, 2026 retirement date for 2.5—that timeline no longer applies. We have amended that article to match Google’s current policy.

October action checklist

1. Confirm production Omni workloads use gemini-omni-1.1-flash (preview endpoint is already off).
2. Evaluate gemini-3.8-live (and Extended Thinking if you need deeper reasoning in live audio) for voice agents.
3. Migrate off gemini-3.1-flash-image before October 29, 2026; target gemini-nano-banana-2.1.
4. If you call Antigravity local tools, update parsers for PascalCase parameters and line-range file edits.
5. Do not schedule production work around Gemini 4 Argon until Google ships general developer access.

Related reading

• September 2026 update (includes corrected 2.5 access limits)
• Omni 1.1 Flash announcement