> ## Content Index
> Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Google Ships Side Project While Gemini 4 Stays in the Lab
- URL: https://wire.fourthweb.ai/google-ships-side-project-while-gemini-4-stays-in-the-lab/
- Published: 2026-09-24T09:04:18.000Z
- Updated: 2026-09-24T09:30:44.000Z
- Description: Google just shipped a text-to-speech model while promising its flagship is "almost ready," which is exactly what a company says when it's still behind.
- Author: Travis Wright
- Tags: AI Agent Economy, AI Agents, AI Infrastructure, OpenAI, Anthropic, Google AI, Microsoft

**Google just shipped a text-to-speech model while promising its flagship is "almost ready," which is exactly what a company says when it's still behind.**

### The Summary

- [Google DeepMind's new chief Koray Kavukcuoglu says Gemini 4 is in refinement and targeting launch "much earlier" than year-end](https://www.theverge.com/tech/999802/google-deepmind-gemini-4-timeline-koray-kavukcuoglu?ref=wire.fourthweb.ai), his first public comments as division leader
- [Google quietly released Gemini 3.8 text-to-speech](https://deepmind.google/blog/say-hello-to-gemini-38-text-to-speech/?ref=wire.fourthweb.ai), a capabilities release while the flagship lags
- The timeline admission confirms Google is playing catchup on frontier models while competitors ship

### The Signal

[Kavukcuoglu's comments to The Information](https://www.theverge.com/tech/999802/google-deepmind-gemini-4-timeline-koray-kavukcuoglu?ref=wire.fourthweb.ai) mark his first media appearance since taking over DeepMind. His language is telling: "as soon as possible" and "we see the results and we are excited" sounds less like confidence and more like internal pressure to ship. Google has been notably absent from the frontier model race this year while [OpenAI](https://wire.fourthweb.ai/tag/openai/), [Anthropic](https://wire.fourthweb.ai/tag/anthropic/), and even smaller labs have pushed out competing releases.

The [Gemini](https://wire.fourthweb.ai/tag/google-ai/) 4 timeline matters because Google invented the transformer architecture that powers every modern LLM, yet keeps getting beat to market by startups and [Microsoft](https://wire.fourthweb.ai/tag/microsoft/)\-backed competitors. Kavukcuoglu says they'll "continue the fast-paced iteration" after an early release, which translates to: we're shipping something that isn't quite done because we can't afford to wait anymore.

> "Our intention is to release an early post-training output because we see the results and we are excited."

Meanwhile, [Google shipped Gemini 3.8 text-to-speech](https://deepmind.google/blog/say-hello-to-gemini-38-text-to-speech/?ref=wire.fourthweb.ai) this week, a capability expansion that:

- Shows Google can still execute on specific modalities
- Keeps the Gemini brand active while the flagship model remains in development
- Signals the company is thinking beyond text, even if the core reasoning model trails competitors

The [Hacker News thread drew 121 comments and 249 points](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/?ref=wire.fourthweb.ai), suggesting developer interest remains high despite Google's lagging release cadence. But interest doesn't equal market position. Every month without a GPT-4 class competitor is another month enterprises build on OpenAI or Anthropic infrastructure.

### The Implication

If you're building agent systems, Google's timeline matters less than their pattern. They have the research talent and [compute](https://wire.fourthweb.ai/tag/ai-infrastructure/) resources to compete, but organizational velocity is the constraint. Don't wait for Gemini 4 to finalize your model strategy. The companies winning in the agent economy right now are model-agnostic, swapping backends as performance and price shift.

For Google, "much earlier than year-end" means they're racing an internal deadline, not the market. The market already moved. Watch whether Gemini 4 actually advances the frontier or just catches up to where others were six months ago. That gap tells you everything about whether Google can lead in Web4 or just follow with better infrastructure.

### Sources

[The Verge AI](https://www.theverge.com/tech/999802/google-deepmind-gemini-4-timeline-koray-kavukcuoglu?ref=wire.fourthweb.ai) | [Hacker News Best](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/?ref=wire.fourthweb.ai) | [Google DeepMind](https://deepmind.google/blog/say-hello-to-gemini-38-text-to-speech/?ref=wire.fourthweb.ai)