Gemini 4 Release Date: What We Know and What's Rumored

Google has not announced a Gemini 4 release date, and as of August 2026 the model exists only as a training run. The only official facts come from two statements in late July: a single line at the end of Google's Gemini 3.6 Flash launch post, and a follow-up from CEO Sundar Pichai on Alphabet's Q2 earnings call. Everything else circulating online, including the widely repeated late-2026 window, is inference from Google's past release cadence rather than anything the company has committed to.

Here is what is actually confirmed, why a real date does not exist yet, and what the timing signals point to.

What Google has confirmed about Gemini 4

Google confirmed Gemini 4 is in pre-training on July 21, 2026. The confirmation arrived at the tail of the post announcing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber, where the Gemini team wrote that it had "already started our most ambitious pre-training run yet, for Gemini 4." Google AI Studio lead Logan Kilpatrick repeated the milestone on X the same day.

Pichai widened the picture two days later on the Q2 2026 earnings call. Asked whether Gemini can stay at the frontier, he said the next generation would need much larger base models and confirmed the model by name: "We are now training Gemini 4, and we're being very ambitious with it." He described the effort as significantly larger than any prior Gemini and named coding and agentic coding as the capability areas Google most needs to improve.

One more signal came out of the same call. Pichai said Google intends to move toward an almost monthly release cadence for Gemini models, a shift from its slower historical rhythm. That is the complete confirmed record. No release date, model ID, benchmark, price, context window, or variant lineup has been published.

Why there is still no Gemini 4 release date

Pre-training is an engineering milestone, not a product. A frontier model has to clear several stages after the training run before anyone outside Google can use it, and none of them moves quickly. Post-training covers instruction tuning, safety alignment, and tool-use integration. Internal evaluation stress-tests reasoning, long-context reliability, and agent behavior. Only then comes limited preview access, and finally general availability with a stable model ID and documentation.

Google's own recent history shows how long that gap can run. The confirmation of a training run tells you the next generation is underway. It does not tell you the model is close, and it is not the same as a model page appearing in the API catalog.

What the timing signals point to

Prediction markets currently put a Gemini 4 launch most likely in late 2026, though the odds fall off sharply before then. On Manifold, the community-run market for a Gemini 4.0 release gives roughly a 74% chance of a launch before November 2026, about 85% before December, and about 89% before the end of the year. Traders price an October arrival at closer to 32%. These are crowd forecasts tied to a defined resolution condition, not insider knowledge, but they track the same late-2026 window that analysts have inferred from Google's six-month development cycles.

Tech press estimates land in the same range. 9to5Google reasoned that, based on past releases, Gemini 4 should arrive in November or December 2026. Google has attached no date to that window, so treat every specific month as an educated guess until a model ID appears in the Gemini API or Vertex AI catalog.

The Gemini 3.5 Pro delay hanging over the timeline

The Gemini 3.5 Pro delay is the reason many developers read the Gemini 4 tease with caution. Google announced the 3.5 series at I/O in May 2026 and said the Pro flagship would follow the next month. It has not shipped. Google now describes 3.5 Pro as testing with partners and says it will launch "as soon as it's ready," and Bloomberg reported that coding performance drove the slip after a late-June training-data update fell short of internal targets.

Promoting the next generation before the current flagship reaches general availability is what unsettled parts of the developer community. It also matters for the Gemini 4 timeline: Gemini 3.5 Pro and Gemini 4 are separate models trained in parallel, so a still-unreleased flagship in the current generation is a live signal about how long a finished-looking model can sit before launch.

Gemini 4 versus Gemma 4

Gemini 4 and Gemma 4 are different products, and the similar names are the biggest source of confusion in search results. Gemini 4 is Google's proprietary frontier model, unreleased and accessible only through Google's infrastructure once it ships. Gemma 4 is the open-weight family Google released in April 2026, built from Gemini research but downloadable and runnable on your own hardware. Benchmark numbers and specs that surface under a "Gemini 4" search usually belong to Gemma 4, not the unreleased frontier model.

What to use until Gemini 4 ships

The practical Google options today are the current-generation Gemini models, all available while Gemini 4 trains. Gemini 3.6 Flash is the current workhorse at $1.50 per million input tokens and $7.50 per million output tokens, with stronger coding and agentic benchmarks than the model it replaced. Gemini 3.7 Flash, launched August 13, 2026, sits alongside it. For heavier reasoning and long-context work, Gemini 3.1 Pro runs $2.00 input and $12.00 output per million tokens on prompts up to 200K tokens. High-volume, latency-sensitive jobs can drop to Gemini 3.5 Flash-Lite at $0.30 and $2.50. Pricing is current as of August 2026 and set on the standard paid tier.

Waiting for an unannounced model carries a real opportunity cost. Building on a current model now, with a replaceable model layer, lets you switch to Gemini 4 the day it lands without rearchitecting anything.

If you would rather build than wait, the model you use should not lock you in. Emergent builds full-stack applications from a plain-language description, and its Universal LLM Key gives you access to Claude, GPT, and Gemini through one credential and unified billing, with no separate API accounts to set up. Emergent supports Google's current Gemini models today, so when Gemini 4 becomes available on the platform, you can start using it without new keys or extra configuration. Start Building on Emergent.