Gemini 4 Is Coming Early: Everything We Know So Far
Google’s next flagship model is closer than expected. At The Information’s AI Agenda Live summit this week, Google DeepMind’s Koray Kavukcuoglu said Gemini 4 has entered post-training and that Google wants to release an early version as soon as possible. Here’s what we know — and what’s still unknown.
What’s in this guide
- What Google said
- The timeline so far
- What Gemini 4 is expected to focus on
- Why Google is in a hurry
- What’s still unknown
- FAQ
What Google Said
Kavukcuoglu, promoted to Senior Vice President of Google DeepMind in August, made his first public appearance in the role this week. He said Google intends “to roll out an early post-training version as soon as possible,” because the team has “already seen promising results.” Reporting on his remarks says Google hopes to ship Gemini 4 much earlier than the end of 2026.
The Timeline So Far
| Date | Milestone |
|---|---|
| July 21, 2026 | Google announces the start of Gemini 4 pre-training |
| August 12, 2026 | Koray Kavukcuoglu promoted to SVP of Google DeepMind |
| September 2026 | Gemini 4 enters post-training, roughly two months after pre-training began |
| Before end of 2026 (target) | Early version release |
No official release date has been announced. The “early version” wording suggests an initial preview, with a broader rollout later.
What Gemini 4 Is Expected to Focus On
Coding
Continuing the push seen in Gemini 3.8 Flash, which made big gains on coding benchmarks.
Autonomous agents
Models that can plan and carry out multi-step tasks with less supervision.
Long-horizon work
Staying reliable across long agentic workflows, not just single answers.
Scale
Described as significantly larger than previous Gemini models, though Google hasn’t shared specifics.
Why Google Is in a Hurry
September has been an intense month. OpenAI launched GPT-6 Astra, Anthropic shipped Claude Fable 5.1, and Meta released Muse Spark 1.3 alongside a hit consumer agent. Google’s recent releases have leaned toward fast, affordable Flash models — Gemini 3.8 Flash is among the best-value models available — while rivals have been setting the pace at the very top of the capability rankings.
Gemini 4 is Google’s chance to compete for the frontier again. It also matters for Google’s products: a stronger model would flow into Search, Android, Workspace, and Google’s growing lineup of agent tools.
What’s Still Unknown
- Release date: only “as soon as possible” and “earlier than year-end.”
- Pricing: no details yet.
- Benchmarks: Google hasn’t shared numbers beyond “promising results.”
- Safety approach: given the cyber capabilities of current frontier models, expect gated access for some features, as with Gemini 3.8 Flash Cyber.
- Which products get it first: the Gemini app, API, or Workspace.
FAQ
When will Gemini 4 be released?
Google hasn’t given a date. It says it wants to release an early version as soon as possible, and well before the end of 2026.
Will Gemini 4 be free?
Unknown. Google typically offers some access in the free Gemini app, with more capable tiers in paid plans, but hasn’t confirmed plans for Gemini 4.
Should I wait for Gemini 4 before choosing an AI tool?
Not necessarily. Use what works for you today — switching between assistants is easy, and there will always be a next model.
Related Reading on FutureLume
- GPT-6 Astra vs Claude Fable 5.1 vs Gemini 3.8 Flash vs Muse Spark 1.3: September 2026’s AI Model Launches Compared
- Best LLMs in 2026: Which One Should You Actually Use
- The AI Data Center Energy Crunch in 2026: What’s Actually Straining the Grid
- The AI Bubble Question in 2026: What the Funding and Revenue Data Actually Shows
Gemini 4 is in its final stages of training, and Google is clearly motivated to ship it fast. Expect a focus on coding and agents, and likely an early preview before a wider launch. Until Google shares benchmarks and a date, treat the “coming early” message as intent, not a promise.
