The New Gemini: Google Releases Gemini 4 Carbon and the Decisions API
Open the Gemini app this week and you may spot a model name in the picker that wasn't there a few days ago. Developer chats are buzzing about a Decisions API, and some coverage mentions model names you won't find on any official page. So we checked Google's own blog, the Gemini API changelog and the Google Cloud press release, and kept only what holds up.
In short: On 10 November 2026 Google made Gemini 4 Carbon, its most capable model so far, available to paying developers and Google AI Ultra subscribers. On 11 November it added Gemini Decision and the Decisions API for fast, probability-based decisions, and Google Cloud launched Gemini for Healthcare in preview for healthcare organizations.
What is the new Gemini? The models Google announced
On 30 September Google announced Gemini 4 Argon, first for trusted cyber defenders through its Fairwind program. Per the official changelog, Argon became generally available in the API on 13 October, the same day it reached the Gemini app for Google AI Pro and Ultra subscribers.
The name causing confusion is Gemini 4 Carbon, and it is official: Google DeepMind published its model card, and Koray Kavukcuoglu, Google's Chief AI Architect, announced on 10 November that it is now available. Google calls it its strongest model for cybersecurity and long-horizon agentic work, where the model works step by step over a long stretch without constant input.
What does Gemini 4 Carbon actually offer?
- Inputs: text, images, audio and video; output is text only.
- Context window: up to 1 million tokens, input and output combined.
- Knowledge cutoff: March 2026, according to the model card.
- Target uses: complex coding and debugging, scientific research, high-precision knowledge work in law and finance, and cyber defense.
- Thinking: it uses more thinking tokens than Argon, with configurable thinking levels.
Interestingly, Google's own post says Argon remains the right choice for most software engineering and knowledge work, while Carbon is meant for problems that test the limits of reasoning, like an obscure bug in a large codebase. Carbon explores more and runs longer verification loops, which costs more time and tokens.
Gemini Decision and the Decisions API: a tool for developers
This got less attention but may matter most to app builders. On 11 November Google's changelog added a model called gemini-decision-1 and the Decisions API. Unlike a chatbot, it never writes text; it answers specific questions with probabilities.
Picture an online shop receiving hundreds of messages a day. You ask: is this a complaint? Which department should get it? How urgent is it, from 1 to 5? You get probabilities and a confidence value back, and your code decides whether to act automatically or hand the case to a person.
- Three question types: yes/no, choice from a list, and score against a rubric.
- Up to 1,000 questions per request against the same input.
- Accepts text, images, audio, video and PDF files.
- Output token count is always zero, because the result is numbers, not text.
- Currently available on the Gemini Developer API under v1beta.
Gemini for Healthcare: a preview for organizations
Also on 10 November, Google Cloud announced Gemini for Healthcare, an agentic product for hospitals and health plans that runs multi-step tasks with human oversight, such as summarizing medical records and reviewing insurance coverage. It is in preview for Gemini Enterprise Standard and Plus customers, and early customers include Stanford School of Medicine and Optum.
Google says patient data stays in the customer's environment and is not used to train models. This is an enterprise product, not a self-diagnosis app, and any clinical use should be reviewed by qualified personnel.
Other announcements from the past two weeks
- Gemini Omni 1.2 Flash: generally available since 5 November for video generation and conversational editing, with clips up to 15 seconds.
- November Pixel Drop: rolling out since 10 November with Gemini Intelligence features for the Pixel 10 series.
- Older models retired: Gemini 2.5 Pro and 2.5 Flash requests are rerouted to newer models on 1 December 2026, so check any code that depends on them.
Quick comparison: what changed?
| Model | When | Who can use it | Best for |
|---|---|---|---|
| Gemini 4 Argon | Generally available since 13 October | AI Pro and Ultra subscribers; developers on the paid tier | Most everyday coding and analysis |
| Gemini 4 Carbon | Available since 10 November | AI Ultra subscribers; paying developers and businesses | The hardest problems and long tasks |
| Gemini Decision 1 | 11 November | Developers via the Gemini API | Fast classification, routing and scoring |
On API pricing, Google's official pricing page says Carbon and Argon are not on the free tier. Their listed rate through 30 November is $2 per million input tokens and $10 per million output tokens, and the rate doubles from 1 December.
How can you try the new Gemini from an Arab country?
The Gemini app supports Arabic, and Google's help page lists most Arab countries, including Egypt, Saudi Arabia, the UAE, Kuwait and Morocco; Syria is not listed. The real limits come from your plan:
- Free: install the Gemini app or open gemini.google.com with your Google account. No Carbon or Argon, but you can chat in Arabic and generate images. For more ideas, see ways to use Gemini Pro without a subscription.
- Google AI Pro: gives you Argon inside the app.
- Google AI Ultra: the only plan that gives you Carbon in the app, in AI Mode in Search and in Antigravity.
- Developers: open Google AI Studio, enable billing to use Carbon, or try the Decisions API on a small project.
Practical use cases worth trying
- Developers: give Carbon a recurring bug with its logs and ask for a root-cause analysis, not a quick patch.
- Support teams: use the Decisions API to sort incoming messages by type and priority.
- Students: upload a long PDF, ask for a chapter-by-chapter comparison, then verify every citation yourself.
- Everyday users: the free version is enough for drafting messages, summarizing articles and trying image generation with Nano Banana.
For free alternatives, see our guide to the best free AI tools of 2026.
FAQ
Is Gemini 4 Carbon free?
No. In the app it requires Google AI Ultra, and the API's free tier doesn't include it.
What is the difference between Gemini 4 Argon and Gemini 4 Carbon?
Carbon is built on Argon but thinks longer. Google recommends Argon for most tasks, Carbon for very complex ones.
Does Gemini work in Arabic?
Yes. Arabic is a supported language in the Gemini app, and the service is available in most Arab countries.
What is the Gemini Decisions API, in one line?
A developer endpoint that answers yes/no, choice or score questions with probabilities, not text.
The takeaway
November brought three real announcements: Carbon for the toughest work, the Decisions API for app builders, and a healthcare product for organizations. Everyday users won't see much change without Ultra; developers should give the Decisions API an hour this week.
Sources: Google's The Keyword blog, Google DeepMind's Gemini 4 Carbon model card, the Gemini API changelog and pricing pages in Google AI Studio, the Decisions API documentation, Gemini app release notes, the Google Cloud press release, Google's Pixel blog, and the Gemini Help Center.
Because AI tools change every week, we keep up with them with help from following Arabic websites and specialized blogs, to what Egypt stores, Kuwait stores, vitamin stores and international sites offer, with Mashhor for social media services as a key to growth.
That's it, friends. We hope you found this useful โ
Best wishes from the #Ezznology team.
Join our community on Telegram and our Facebook group, and follow us on Google News.
You may also like:
Gemini 3.1 Pro for Free: 5 Smart Ways Without a Paid Plan
Gemini Live Finally Gains a Memory Feature
