Gemini 4 Pro is still not an official model name
What went GA on 2026-10-13 is Gemini 4 Argon (gemini-4-argon); the official catalog has no id called gemini-4 or gemini-4-pro
As of 2026-10-19, Google's official Gemini API documentation has no model called Gemini 4 Pro. The one model in the Gemini 4 family that is live is Gemini 4 Argon (id gemini-4-argon), generally available since 2026-10-13; Google calls it its most capable model built for complex coding, agentic, math and science tasks, and the model page states a unified 1M context window with long decodes up to 1M tokens. The earlier official record still holds too: the official blog of 2026-07-21 said the Gemini 4 pre-training run had started and the same post states Gemini 3.5 Pro is still testing with partners, and the API model catalog last updated 2026-09-24, re-checked 2026-09-29, had neither a gemini-4 nor a gemini-4-pro id, price or model card.
Updated 2026-10-22
One key across Claude, GPT and Chinese models, billed per token.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
What Google actually said
The live 4-family model
The official release notes mark Gemini 4 Argon (gemini-4-argon) as generally available on 2026-10-13, with the status Released; in the official catalog dated 2026-09-24 there was neither a gemini-4 nor a gemini-4-pro.
Pre-training started
Official blog: the Gemini 4 pre-training run has started. That is not a release and there is no SKU; the release itself came on 2026-10-13.
The live Pro is unchanged
The Pro tier listed as live in the official catalog is still Gemini 3.1 Pro, id gemini-3.1-pro-preview; the release notes, the model page and the deprecation table do not carry the name Gemini 4 Pro.
In catalog
The official catalog lists gemini-3.8-flash and gemini-3.8-live — the 3.x line keeps growing, which is a separate fact from Gemini 4; the shutdown date in the deprecation table for gemini-3.1-pro-preview reads No shutdown date announced.
Why people type Gemini 4 Pro
Three things to separate when you search Gemini 4 Pro: the first officially confirmed fact is that pre-training started on 2026-07-21, and the same post says Gemini 3.5 Pro is still testing with partners; the live Pro in the official catalog is gemini-3.1-pro-preview, alongside gemini-3.8-flash and gemini-3.8-live; and the model that went GA in the Gemini 4 family on 2026-10-13 is Gemini 4 Argon (gemini-4-argon), while the official documentation has never carried a model called Gemini 4 Pro. Gemini 4 is not a rename of 3.5 Pro.
The competitive context
Re-checked 2026-09-18 against the catalog dated 2026-09-17: what was added is gemini-3.8-flash and gemini-3.8-live, not Gemini 4, and the only official statement about Gemini 4 was still the 2026-07-21 pre-training line. On 2026-10-08, Google marked gemini-3.7-flash as deprecated: it is no longer available to select in Google AI Studio, and API requests to it are automatically routed to gemini-3.8-flash. On 2026-10-13, the Gemini 4 family went GA under the name Gemini 4 Argon.
Public crumbs
2026-07-21 Google's official blog confirms the Gemini 4 pre-training run has started and that Gemini 3.5 Pro is still testing with partners.
2026-09-17 (catalog marked "Last updated 2026-09-17 UTC"): the API catalog added gemini-3.8-flash and gemini-3.8-live, and still has no gemini-4*.
2026-10-19: this page was re-checked against the official release notes, the model page and the deprecation table, and the official documentation carries no model called Gemini 4 Pro; in the same window Google marked gemini-3.7-flash as deprecated on 2026-10-08, with API requests to it routed automatically to gemini-3.8-flash, and on 2026-10-13 the Gemini 4 family went GA as Gemini 4 Argon; the catalog dated 2026-09-24, re-checked 2026-09-29, had no gemini-4 and no gemini-4-pro id either.
Confirmed vs rumor
Confirmed
Confirmed, each with a first-hand source: the Gemini 4 pre-training run started 2026-07-21, and the same post states Gemini 3.5 Pro is still testing with partners; in the API catalog dated 2026-09-24 and re-checked 2026-09-29 the live Pro is gemini-3.1-pro-preview, alongside gemini-3.8-flash and gemini-3.8-live; on 2026-10-08 Google marked gemini-3.7-flash as deprecated, with API requests routed automatically to gemini-3.8-flash; and on 2026-10-13 Gemini 4 Argon (gemini-4-argon) went generally available, marked Released in the release notes, which Google calls its most capable model for complex coding, agentic, math and science tasks, with a unified 1M context window and long decodes up to 1M tokens on the model page.
Rumor
"Gemini 4 Pro ships in October" and "the first checkpoint leaked" have no source in the official documentation: beyond the catalog dated 2026-09-17 and the 2026-07-21 blog post there is no official date, id, price or benchmark, and leaked third-party scores are never mixed with official numbers. What went GA on 2026-10-13 is Gemini 4 Argon (gemini-4-argon); the official documentation still carries no model called Gemini 4 Pro.
What to use instead
Need Google stack today
The ids listed in the Google catalog today: gemini-4-argon (the 4 family), gemini-3.1-pro-preview, gemini-3.8-flash and gemini-3.8-live, all straight from the official catalog.
Need 2026 frontier coding
gemini-4-pro is not in the official catalog, so a config that sets it gets no response.
How this page flips green
The model id Google ends up publishing may be gemini-4, gemini-4-pro or something else — the name in the official documentation today is Argon, not Pro. Benchmarks and context limits follow the official model card: the Gemini 4 family model card is in Google's Gemini API documentation (ai.google.dev), where the live model Gemini 4 Argon (gemini-4-argon) went GA on 2026-10-13 with a unified 1M context window and long decodes up to 1M tokens.
On QCode
QCode does not offer the Gemini series for now; models on QCode include claude-sonnet-5-5, gpt-6.1-sol and glm-5.3, billed per token, with per-model rates on /models. Want to use the models QCode already offers for just a few days? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days (top-ups start at $5).
Gemini 4 Pro FAQ
Is Gemini 4 Pro released?
No. Per the catalog dated 2026-09-24 and re-checked 2026-09-29, there is no gemini-4-pro id, and the official blog post of 2026-07-21 only describes the pre-training that started that day. As of the 2026-10-19 re-check, the model in the Gemini 4 family that went GA on 2026-10-13 is Gemini 4 Argon (id gemini-4-argon) in the official release notes and on the model page, and the official documentation carries no name Gemini 4 Pro.
Did Google announce Gemini 4?
It was announced, but not under the name Pro. The official blog of 2026-07-21 said the Gemini 4 pre-training run had started; the release notes of 2026-10-13 mark Gemini 4 Argon (gemini-4-argon) as generally available, and the banner at the top of Google's docs says Gemini 4 Argon is now available. Google calls Gemini 4 Argon its most capable model, built for complex coding, agentic, math and science tasks.
Is Gemini 4 just 3.5 Pro renamed?
It is not a rename. Gemini 3.5 Pro is still the 3.x SKU described as "testing with partners" on 2026-07-21, and Gemini 4 is the next pre-training run mentioned in the same post; what went GA on 2026-10-13 is the new model name Gemini 4 Argon (id gemini-4-argon), and there is no renaming note in the official record.
When will Gemini 4 Pro launch?
There is no official date. The catalog's own stamp reads Last updated 2026-09-24 UTC (re-checked 2026-09-29) and contains no Gemini 4 row; in the official deprecation table the Gemini 3 rows, for example gemini-3.1-pro-preview, read No shutdown date announced, and the same page says the shutdown dates in that table are the earliest possible dates on which a model might be retired, not fixed dates. Any specific month is not official.
What Gemini can I use for coding now?
Per the catalog dated 2026-09-24 (re-checked 2026-09-29) the coding-side choices are gemini-3.1-pro-preview, gemini-3.8-flash and gemini-3.8-live; read each model card for actual capability. Since 2026-10-13, the only live model in the Gemini 4 family on the official model page is Gemini 4 Argon (a unified 1M context window and long decodes up to 1M tokens); the Gemini series is not currently offered on QCode, and for coding you can pick claude-sonnet-5-5, gpt-6.1-sol and glm-5.3.
Will QCode list Gemini 4 Pro on day one?
No. QCode does not offer the Gemini series for now, and there is no model called Gemini 4 Pro in the official docs. Models you can call on QCode include claude-sonnet-5-5, gpt-6.1-sol and glm-5.3, billed per token, with per-model rates on /models.
Sources
Google's official blog of 2026-07-21 (Gemini 4 pre-training; 3.5 Pro in partner testing, captured 2026-09-18) and the Gemini API model catalog on ai.google.dev (its own stamp reads Last updated 2026-09-24 UTC, captured 2026-09-29); on 2026-10-19 three pages of the official Gemini API documentation were fetched again: the release notes (the 2026-10-13 Gemini 4 Argon GA entry), the model page (the Gemini 4 Argon entry and its 1M context description) and the deprecation table (the Gemini 3 family rows and the note on shutdown dates).
Start with the models QCode already has
QCode does not offer the Gemini series for now; one key calls Claude, GPT and Chinese models such as DeepSeek, billed per token, with per-model rates on /models.
Related
AI Model Radar 2026
Includes the Gemini 4 rumored card.
How to buy Claude API credits
Console prepaid credits, auto-reload, payment methods, one-year expiry and Max/Team monthly credits, compared with QCode top-ups.
DeepSeek V4 Pro 0813
Open-weight flagship from the same week.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
Not affiliated with Google / DeepMind. Gemini 4 Pro is a search phrase, and the official documentation has no model with that name; what Google released on 2026-10-13 is Gemini 4 Argon. The official docs support these points: model names, release dates, Google's own specifications and the shutdown dates in the deprecation table. Community rumours and leaked benchmarks are not release evidence.