Gemini 4.1 "Lithium": There Is No Leak, No Codename, and One Very Good Joke

There is no Gem​ini 4.1. That is the finding, and it took about four minutes to establish — the name we were handed, lifted from a post that reads "Gem​ini 4.1 Lithium when", appears on no G​oogle surface that ships anything to a developer. G​oogle's frontier model, Gem​ini 4 Argon, was announced on September 30 and still cannot be called by anyone outside a vetted cyber-defence programme; the top of the line you can actually call today is Gemini 3.8 Flash, and it is a 3.x model. What exists as of October 1, 2026 is a chemical-element naming convention introduced one day earlier, a joke about naming an agentic model after an unreactive gas, and a question mark. This piece is about the question mark: what the post is, what it is not, and what would have to appear before "Gem​ini 4.1" becomes a string you can write code against.

Where the word "Lithium" comes from

The signal is one X post, published 2026-10-01 at 00:50 UTC by the account @teortaxesTex, reading in full: "Gem​ini 4.1 Lithium when". At the time of writing it carries 35 likes, 0 retweets, 1 reply, 3 bookmarks and roughly 2,466 views. It is a quote post, and the quoted post is the interesting half.

That quoted post is from @bryancsk, published 2026-09-30 at 20:41 UTC — about four hours earlier, and hours after G​oogle's own Argon announcement landed that same day. It reads: "Naming an agentic model after an unreactive element, and a word that also means 'lazy' in Greek is a surprising choice". Both halves are accurate about the word argon: it is a noble gas that forms essentially no compounds, and ἀργόν, the Greek neuter of ἀργός, means idle or lazy. So the complaint is a naming joke about Gem​ini 4 Argon, and the "Gem​ini 4.1 Lithium when" reply is the punchline — if the Gem​ini 4 family is going to be named after elements, the next one is obviously the next element, and the "when" is impatience, not information.

That distinction is the whole article. A leak post asserts that something exists and is being hidden. This post asserts that something does not exist yet and is wanted. There is no document behind it, no staff account, no repository, no arena alias, no filing. Anyone who tells you a Gem​ini 4.1 codename leaked this week has read the same one-liner and inferred a leak from its shape.

Does "Lithium" have any history as a G​oogle codename?

Partly, and the history is a trap. In mid-October 2025, two anonymously labelled models appeared in live matchups on the public arena and did not appear on its leaderboard. Community trackers named them lithiumflow and orionmist and mapped them — as analysis, never as confirmation — onto what people expected to become the Gem​ini 3 line, with lithiumflow as the stronger variant. Several outlets covered it on 2025-10-20 under a rumoured announcement date of October 22. G​oogle never confirmed the mapping, and nothing about it survived into the naming G​oogle actually uses, which is tier labels: Flash, Flash-Lite, Pro, Live, Transcribe.

Two things follow. First, "lithiumflow" and "Lithium" are different strings; a 2025 temporary arena label does not make a 2026 element name real. Second, the only element-style name G​oogle has shipped is the one it announced on September 30, 2026 — Argon, which coverage on the day described as the first Gem​ini with a codename-style name, in the manner of O​penAI's GPT-6 Astra and Sol. One data point is not a convention. It is a guess that a convention might start.

Four checks, run on October 1, 2026

Everything below was read directly today, not recalled, because "no such model" is a claim that goes stale in days.

• Gem​ini API model list — the highest generation on the page is the 3.8 family (Flash, Live, Live Extended Thinking, Flash TTS, Flash-Lite TTS) plus Gem​ini 3.7 Flash, 3.6 Flash, 3.5 Flash, 3.1 Pro and 2.5 Pro. No Gem​ini 4, no Gem​ini 4.1, no Lithium. Page footer dated 2026-10-01 UTC.

• Gem​ini API changelog — newest entries are September 22, 2026 (Gem​ini 3.8 Flash TTS and Flash-Lite TTS to GA, plus a voices endpoint), September 18 (Gem​ini 2.5 access limited to prior users), September 17 (Antigravity Agent preview rollover), September 15 (Gem​ini 3.8 Live GA), September 2 (Gem​ini 3.8 Flash GA). Page last updated 2026-09-23. G​oogle's own developer surface has not been touched by the Argon announcement, let alone by a 4.1.

• The public arena text leaderboard — the only G​oogle entry using an element name is gemini-4-argon-high, listed at rank 1 with a score of 1525 ±9, marked preliminary, on 4,942 votes, priced $2/$10 with a 1M context window. No entry on the board contains "lithium". No G​oogle entry contains "4.1" — the strings that look close (gemini-3.1-pro-preview, gemini-3.1-flash-lite-preview) are 3.1, which is a different thing and a year of drift away from being the same thing.

• The catalogue we sell from — google/gemini-4-argon, google/gemini-4 and google/gemini-4.1 all return "model not found" as of today. The G​oogle models that resolve are 3.8 Flash and 3.6 Flash at $0.75 per million input and $3.75 per million output, 3.5 Flash at $1.50/$9, 3.1 Pro Preview at $2/$12 with a $4/$18 tier past 200,000 tokens, and 2.5 Pro at $2.50/$15. Argon is not on our platform and neither is anything called 4.1, and we are not going to imply otherwise for the sake of a sentence.

What G​oogle has actually put on the record

To be plain about the thing this post is reacting to, because it is easy to lose in the joke: the model itself is from September 30, 2026, and this is not a write-up of its launch. G​oogle's Argon announcement is first-party and specific about a few things and silent about the ones that matter for planning. It is the first model of the Gem​ini 4 generation by G​oogle's own framing. Access goes first to vetted cyber defenders through the Fairwind Program, then to paid API customers and AI Ultra subscribers, then to developers, enterprises and consumers, with no dates attached to any of those later stages. Pricing is introductory — $2 per million input tokens and $10 per million output, cached input at 95% off, stepping to $4 and $20 after an introductory period G​oogle did not define. The output limit is 1 million tokens, up from 64K.

What is not on the record is the part that matters for a question like "when is 4.1": there is no Gem​ini API model ID for Argon, no model card, no knowledge cutoff, no published context window and no general-availability date. A version increment cannot be scheduled against a version that has not shipped, which is the structural reason the "4.1 when" question has no answer yet — not a reason someone is sitting on one.

The cadence argument deserves the same scepticism. The Flash line has genuinely been fast: 3.5 Flash, then 3.6 Flash on July 21, 3.7 Flash on August 13, 3.8 Flash to GA on September 2. That is a train running every three to six weeks. But the frontier line is a different train, and the counter-example is sitting right there in G​oogle's own recent history: Gem​ini 3.5 Pro was reported as repeatedly postponed internally and never released at all. A numbering plan is not a shipping plan.

What would turn "Gem​ini 4.1" into a real story

Every one of these is checkable by you, and each has a precedent from this year. When two or more line up, the speculation has become news.

• A model ID in the Gem​ini API model list — this is the first hard artefact. Argon does not have one yet; whatever comes next will.

• A changelog entry with a date. G​oogle's changelog is the developer surface's diary, and it has not been touched since September 23.

• A model card with a knowledge cutoff and a context window. Its absence is currently the cleanest way to tell an announcement from a release.

• An arena alias that accumulates votes rather than appearing once. The Argon entry carries 4,942 votes and a preliminary flag — a used, disputed, argued-over alias. A single sighting with no votes is a test, not a model.

• A named successor from G​oogle itself. The company has said nothing about what other Gem​ini 4 models will be called or when they ship, and until it does, "Lithium" is a chemistry joke with a follower count.

What to run while the question is open

If you came here because you want a next-generation G​oogle model in production this week, the honest answer is that the frontier is gated and the increment does not exist, so the useful question is what the best callable option is right now. The answer from our own catalogue is the 3.8 line: Gem​ini 3.8 Flash is on OrcaRouter at G​oogle's list price, $0.75 per million input tokens and $3.75 per million output with cache reads at a tenth of that, a 1M-token context window and a 65K output ceiling. Gem​ini 3.6 Flash is the other same-tier option if you want a second data point, Gem​ini 3.5 Flash runs $1.50/$9, and Gem​ini 3.1 Pro Preview runs $2/$12 for the reasoning-heavy work with a $4/$18 tier past 200,000 tokens.

The reason this matters more than usual in a week when G​oogle moves a number is the pricing mechanics. We pass provider list price through at 0% markup, so when G​oogle changes a G​oogle rate, the rate on this side changes the same day rather than at the next invoice — which is the practical hedge against an introductory period that G​oogle has not defined an end for. And if you want to trial something unproven the moment it appears, automatic failover across providers means an untested model can sit behind a proven one instead of holding your production path hostage. One API for 200+ models, one key, no second contract, nothing to re-integrate when the version number finally does move.

Questions the post leaves open

Does the arena entry for Gem​ini 4 Argon mean a wider Gem​ini 4 rollout is close?

No, and this is the most common misreading of the week. An arena entry means a configuration was evaluated on a public harness and that people argued about it in votes; it says nothing about API access, eligibility, dates, or what the next version in the family will be called. The Argon entry is marked preliminary, and the model still has no API model ID. Treat the arena as evidence that a model exists, never as evidence about when you can use it.

Is Gem​ini 4.1 the same thing as the Gem​ini 3.5 Pro that never shipped?

There is no evidence for that, and it is worth saying why the rumour is attractive rather than pretending it is not. Gem​ini 3.5 Pro was reported as repeatedly postponed internally and never released, which leaves an obvious gap in the Pro line where a new name could plausibly land. But G​oogle's own framing calls Argon the first model of the Gem​ini 4 generation, which makes the 4.x line a successor generation rather than a rebadged cancellation, and no first-party material connects the two.

When should anyone start planning for a Gem​ini 4.1?

When a model ID appears in the API model list. Everything before that point is scheduling guesswork with a benchmark culture attached, and the cost of planning early is real: effort spent re-integration-testing against a version number is effort not spent on the model you can call today.

The only genuinely new fact produced on October 1, 2026 is that somebody wants the next Gem​ini badly enough to post about it — and that is worth exactly what a wish is worth. The durable content of the week is the naming turn: a frontier lab that spent years shipping tier labels has started shipping elements, and the first element it picked invites a joke about laziness that will follow the model for its whole life. Whether the second element is lithium is a chemistry guess until G​oogle says otherwise, and the string that will settle it is not a codename in a post. It is a line in a changelog. Watch the developer surface, and until it moves, the model to build against is the one that answers today.