
Gemini 4.0 or Gemini 4? Why the Dotted Spelling Is the Same Model
- OrcaNEWOrca: OrcaCyber Zero 1.52026-10-10$3.00 / $7.50 per 1M tokens · 56 tok/s
- openaiOpenAI: GPT-6.1 Sol2026-09-2952Intelligence
- anthropicAnthropic: Claude Sonnet 5.52026-09-2856Intelligence
- typesafeTypeSafe: Jev 1.132026-09-24$0.04 / $0.00 per 1M tokens · 993 tok/s
- OpenAIOpenAI: GPT-6 Luna2026-09-2238Intelligence
- OpenAIOpenAI: GPT-6 Sol2026-09-2248Intelligence
- AnthropicAnthropic: Claude Opus 5.52026-09-2258Intelligence
- xAIGrok 4.72026-09-2146Intelligence
- OrcaOrca: OrcaCyber Zero 1.02026-09-17$3.00 / $7.50 per 1M tokens · 40 tok/s
- OrcaOrca: OrcaVerify Text 1.02026-09-16$2.00 / $0.00 per 1M tokens · 2230 tok/s
- DeepSeekDeepSeek: DeepSeek V4.1 Flash2026-09-1040Intelligence
- OpenAIOpenAI: GPT-6 Astra2026-09-0453Intelligence77Coding
- GoogleGoogle: Gemini 3.8 Flash2026-09-0241Intelligence76Coding
- AlibabaQwen: Qwen3.8 Max (0902)2026-09-0245Intelligence76Coding
- AnthropicAnthropic: Claude Fable 5.12026-09-0153Intelligence82Coding
- TencentTencent: Hy4 preview2026-08-28$0.83 / $2.50 per 1M tokens · 57 tok/s
- AlibabaQwen: Qwen3.8 Flash2026-08-26$0.15 / $0.47 per 1M tokens · 125 tok/s
- z-aiZ.ai: GLM 5.3 Flash2026-08-2642Intelligence72Coding
- DeepSeekDeepSeek: DeepSeek V4 Flash Vision (Exp)2026-08-21$0.22 / $0.66 per 1M tokens · 206 tok/s
- z-aiZ.ai: GLM 5.32026-08-1845Intelligence75Coding
Google's flagship is called Gemini 4 Argon, and the generation it belongs to is called Gemini 4. Google has not written Gemini 4.0 anywhere — not on its model page, not in the announcement post, and not in the one identifier its API publishes for the model, which is gemini-4-argon and carries no dot. So the dotted form is not a second model, not a minor release and not a version bump inside the line. A reader who types it arrives at exactly the same thing a reader who types the plain form arrives at. The dot comes from somewhere real, though: Google shipped a Gemini 2.0 Flash line and an image model it sells as Imagen 4.0, and both of those carry the .0 in the name the company sells them under. Anyone who learned Gemini's version numbers from that generation, or who has met the imagen-4.0-generate endpoint, has a reasonable expectation that the numbering continued. It did not. If what you want is the model itself rather than its name, the page to read is our Gemini 4 Argon explainer; this page is only about the number.
Two spellings, one subject — and how you can tell
The honest way to settle a spelling question is not to argue about it but to measure it, and the measurement points one way. When our query data is collapsed by flattening out spaces and dots, the dotted term and the plain term land in the same bucket: every query that reads gemini 4.0 resolves onto the same subject as the queries that read gemini 4. When the two are kept separate, they do not overlap at all. That combination — identical when you normalise, disjoint when you do not — is the signature of one subject written two ways, not of two subjects that happen to look alike. Across our own Google Search Console window ending 2026-10-12, the dotted family took 2,915 impressions and 57 clicks over 15 queries. The bare term gemini 4.0 took 1,006 impressions and 34 clicks over 4 queries, which is a better click rate than most of the stem manages. This is a spelling people are looking for on purpose, not a typo tail. It is worth saying what the dotted family is not, because that absence is itself a finding: unlike the Argon stem, where misspellings such as agron, argom and aragon make up a measurable tail of their own, the dotted spelling has no misspelling tail at all. Conventions have followers. Typos have variations of themselves. The dotted form has the first and none of the second.
What the dotted form is not
Three readings are wrong, and each is worth one line because each is a thing a reader might reasonably have assumed.
• It is not a different model. There is no second Gemini 4 that the plain spelling does not reach. The saturation on Google's own material is total: the model page, the announcement post and the API model list agree on the name, and the name does not contain a dot.
• It is not a minor release, the way a point release sits under a major version in ordinary software numbering. Google's published 4-series identifier set has exactly one member, and it is gemini-4-argon. There is no gemini-4, there is no gemini-4.0, and there is no gemini-4.1 — the last of those has been checked separately on this blog and the answer did not change.
• It is not a version bump that Google quietly applied and then documented under a different form. If the company had shipped a 4.0 and then renamed it, the identifier would have moved with the name, the way identifiers always do in the Gemini API's model list. It has one entry for the generation, and the entry is Argon.
None of which makes the dotted spelling ignorant. It makes it early. The .0 form was correct for the previous generation in exactly the place readers learned it, and conventions outlive the versions that taught them.
Where the dot actually comes from
This is the part the demand rows make legible, so it is worth being precise about the two precedents rather than gesturing at them.

The first precedent is the Gemini 2.0 line. Google's own model documentation lists Gemini 2.0 Flash and Gemini 2.0 Flash-Lite under its previous-models heading, with the endpoints gemini-2.0-flash and gemini-2.0-flash-lite. That is the only major Gemini line the page writes with a .0. Every other generation it lists is numbered the way Gemini 4 is: 3.8, 3.6, 3.5, 3.1, 2.5. So the dotted form is not a rule readers invented. It is a rule Google applied once, to the 2.0 generation, and then stopped applying.
The second precedent sits outside Gemini entirely but inside the same API surface, which is what makes it sticky. Google's generative-media model list includes imagen-4.0-generate. A developer who has that string in a config file has, quite reasonably, a memory of a versioned Google model whose number is written 4.0. Put the two memories together and the dotted form stops looking like a mistake and starts looking like something you are supposed to be able to type.
You can type it. It just does not resolve to a dotted name, because there is no dotted name to resolve to.
The one 4-series identifier Google publishes
If you are writing code rather than reading prose, the answer is shorter than this page: the string is gemini-4-argon. It is listed as Stable, it is described on the same page as a reasoning model with a unified 1M context window, and it is the only member of the generation that appears anywhere in the model list. That is the identifier to put in a request, and no dotted variant of it exists.
What the model is — what Google attached to the name, how it is positioned, and what was and was not published alongside it — is a separate question with its own page, and this one deliberately does not re-answer it. Our Gemini 4 Argon explainer carries that. If your actual question is which rung of the line Argon occupies and why no Ultra appears beside it, that reading lives in our Gemini 4 Argon tier positioning piece.
Where each dotted query goes next
Three qualified spellings make up most of the dotted family, and none of them is a naming question. Each has a different home, and the useful thing this page can do is point rather than duplicate.
• gemini 4.0 argon — 1,509 impressions and 12 clicks, the largest single row in the dotted family. That reader already has the right name and wants to know what is behind it. The Gemini 4 Argon explainer is the page for that, and the naming question does not need to be walked through there.
• gemini 4.0 pro — 154 impressions and 9 clicks. This is a tier question wearing a spelling, not the other way round. It belongs to the page that defines what a Pro rung is in a line Google announced as a single model; it is not answered here, and answering it here would only put two of our own pages in front of the same reader.
• gemini 4.0 release date — 165 impressions and 2 clicks. The timeline reading has its own page, and using this one to restate it would be the second time we told that story rather than the first time we answered this one.
The pattern across those three rows is the argument for this page existing at all. A reader who types the dotted form is not asking a different question about the model. They are asking about the number, and until now nothing we published answered that in one screen.
What you will not find here
OrcaRouter serves no model id matching gemini-4 under any spelling, in either direction — not the plain form, not the dotted form, not the identifier Google actually publishes. We checked our own live model list on 2026-10-14 and there is nothing on that stem to route, so there is no price to quote and no endpoint to describe. What we do serve from the same vendor is the shipping workhorse generation: google/gemini-3.8-flash and the rest of the 3.x line, all reachable through one API across more than 200 models with no markup on the provider's list price, so a vendor price change reaches you the same day rather than on our schedule. That is a statement about the models we carry, not about the one this page is named after.

It is worth noting what that independent page demonstrates about the naming question, because it is evidence rather than decoration. A third party that measures the model writes the name the same way Google does: Gemini 4 Argon, undotted. Across the vendor's own material, the vendor's own API list and a measurement house with no interest in Google's branding, there is one form and no dotted variant of it. That is as close to a tiebreak as a spelling question gets.
The convention to use going forward
Write Gemini 4 for the generation and Gemini 4 Argon for the model, and put gemini-4-argon in anything that has to execute. If you are the sort of reader who types the dotted form out of habit, nothing breaks — you have been landing on the right model the whole time. What you should stop doing is assuming the dot carries information, because in this generation it does not. It does not mark a point release, it does not mark a smaller variant, and it does not mark anything Google has shipped.
Both spellings will keep appearing in search results, and for two reasons that have nothing to do with anyone being careless. The first is the precedent: gemini-2.0-flash is still a live string in the previous-models list and imagen-4.0-generate is still a live endpoint, so the dotted idiom is still current in the same documentation set, one page away. The second is how search engines treat punctuation, which is loosely enough that a page written either way can surface for either query. The result is that the dotted query keeps being asked, and the useful answer to it is the one this page gives: same model, no dot in the name, and here is where each of the qualified versions actually goes.

One closing note on sourcing, since this page is mostly a set of negatives. Every claim about which form Google uses was checked on 2026-10-14 against Google's own material: the model page, the announcement post and the API model list, which is where the single 4-series identifier comes from. The claim about the 2.0 line and about imagen-4.0-generate comes from the same documentation set, which is what makes it a statement about Google's conventions rather than about this one model. The demand figures are our own first-party Search Console impressions and clicks, and they describe what people type and nothing else. And the availability line was checked against our live model list the same day rather than carried over from an earlier draft, because a confident availability claim that has gone stale is worse for you than no claim at all.
A naming question is also a good moment to check what has gone wrong before: the documented AI incidents and agent failures around models of this class are collected in the OrcaRouter AI Incident Archive , where each entry carries its own sources, severity, confidence grade and disputed flag.
