Google's Gemini models are proprietary large language models offered through the Gemini API, Google AI Studio, the Gemini Enterprise Agent Platform on Google Cloud and the Gemini app, and Google released three Gemini 4 models between October and November 2026. This hub lists the Gemini models in Google's public documentation as of 15 November 2026, with API IDs, release dates, context windows, prices, availability, access restrictions and retirement dates. It is written for developers, buyers, architects and anyone who needs to know which Gemini model is which. Models are listed by release date, newest first, an objective order that implies no ranking. Every fact comes from Google's own documentation, pricing pages, release notes or official blog, cited in the sources list.
Last checked: 2026-11-15.
This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider. It uses Google's public Gemini documentation as of November 2026. Model details change, so confirm against Google's current documentation.
Key takeaways
- Three Gemini 4 models are generally available. Gemini 4 Argon (13 October 2026), Gemini 4 Carbon (10 November 2026) and Gemini 4 Hydrogen (12 November 2026) are listed as GA on Google Cloud's model versions page.
- Every Gemini 4 model has a 1,048,576-token context. Google documents a unified input and output window, with long decode allowing output up to the full window.
- Gemini 4 Argon and Carbon prices double on 1 December 2026. Both are $2 input and $10 output per million tokens until 30 November 2026, then $4 and $20; Gemini 3.8 and 3.6 Flash prices double on 1 January 2027.
- Some variants are restricted. Carbon is offered on Google Cloud to verified enterprise customers in select regions, and the cyber permissive variants of Argon and Carbon are allowlist-only through the Fairwind Program.
- Older models are being rerouted and retired. Gemini 3.7 Flash and 3.5 Flash were deprecated on 8 October 2026, Gemini 3.6 Flash retires on Google Cloud on 19 November 2026, and Gemini 2.5 models are limited to past users on the Gemini API.
What is Google Gemini?
Gemini is Google's family of multimodal large language models. Google documents the models in two places: the Gemini API and Google AI Studio documentation for developers, and the Gemini Enterprise Agent Platform documentation on Google Cloud, which publishes release and retirement dates for each model version. Current generations are Gemini 4, Gemini 3.x and the legacy Gemini 2.5 line.
All Gemini models are proprietary: Google publishes no Gemini weights and states no Gemini parameter counts. Google states that generally available models on Google Cloud are supported for at least 12 months, and that retirement timelines may be extended but will not be moved earlier than the listed date. Facts are as of 15 November 2026.
Current Gemini models at a glance
As of 2026-11-15. Source: Google AI Studio models page, Google Cloud model versions page and Gemini API changelog. Models are listed by release date, newest first. Category is the Gemini generation and tier. On this page, "Not stated by Google" means not stated by Google on the pages checked (2026-11-15).
| Model | API ID or version | Category | Context | Licence | Status |
|---|---|---|---|---|---|
| Gemini 4 Hydrogen | gemini-4-hydrogen | Gemini 4 | 1M (1,048,576) | Proprietary | GA |
| Gemini 4 Carbon | gemini-4-carbon | Gemini 4 | 1M (1,048,576) | Proprietary | GA, verified enterprise customers on Google Cloud |
| Gemini Omni Flash | gemini-omni-1.2-flash, gemini-omni-1.1-flash | Omni (video) | Not stated by Google | Proprietary | GA |
| Gemini 4 Argon | gemini-4-argon | Gemini 4 | 1M (1,048,576) | Proprietary | GA |
| Gemini 3.8 Flash | gemini-3.8-flash | Gemini 3.x Flash | 1M (1,048,576) | Proprietary | GA (Stable) |
| Gemini 3.6 Flash | gemini-3.6-flash | Gemini 3.x Flash | Not stated by Google | Proprietary | Stable on Gemini API; Google Cloud retirement 2026-11-19 |
| Gemini 3.5 Flash-Lite | gemini-3.5-flash-lite | Gemini 3.x Flash-Lite | 1M (1,048,576) | Proprietary | GA (Stable) |
| Gemini 3.1 Flash-Lite | gemini-3.1-flash-lite | Gemini 3.x Flash-Lite | Not stated by Google | Proprietary | GA (Stable) |
| Gemini 3.1 Pro | gemini-3.1-pro-preview | Gemini 3.x Pro | Not stated by Google | Proprietary | Preview |
| Gemini 3 Flash | gemini-3-flash-preview | Gemini 3.x Flash | Not stated by Google | Proprietary | Preview |
Gemini 3.8 Live, 3.8 Live Extended Thinking, 3.8 Flash TTS, 3.8 Flash-Lite TTS and 3.5 Transcribe are also listed as current on the AI Studio models page. No Gemini model is published with open weights. See Open-Weight vs Closed Models Explained.
Key facts for the main Gemini models
As of 2026-11-15. Identical fields for every model.
| Model | Released | Max output | Knowledge cutoff | Input / output | Open weights | Parameters | Positioning (provider wording) |
|---|---|---|---|---|---|---|---|
| Gemini 4 Hydrogen | 2026-11-12 | 65,536 default; up to 1,048,576 with long decode | March 2026 | Text, image, video, audio, PDF / text | No | Not stated by Google | Built for speed and efficiency across everyday coding and agentic workflows |
| Gemini 4 Carbon | 2026-11-10 | 1,048,576 (unified); 65,536 streamed or 32,768 non-streamed per segment | March 2026 | Text, image, audio, video / text | No | Not stated by Google | Frontier model for deep work in complex, real-world tasks; for the most demanding tasks (blog) |
| Gemini 4 Argon | 2026-10-13 | 1,048,576 (unified); 65,536 streamed or 32,768 non-streamed per segment | Not stated by Google | Text, image, audio, video, PDF / text | No | Not stated by Google | Frontier model for deep work; recommended for most software engineering and knowledge work (blog) |
| Gemini 3.8 Flash | 2026-09-02 | 65,536 | Not stated by Google | Text, image, video, audio, PDF / text | No | Not stated by Google | Most intelligent Flash model, for long-horizon software engineering and agents |
| Gemini 3.5 Flash-Lite | 2026-07-21 | 65,536 | Not stated by Google | Text, image, video, audio, PDF / text | No | Not stated by Google | Low-latency, cost-effective model for high-volume agentic workflows |
| Gemini 3.1 Flash-Lite | 2026-05-07 | Not stated by Google | Not stated by Google | Text, image, video, audio (priced) / text | No | Not stated by Google | Not stated by Google |
| Gemini 3.1 Pro | 2026-02-19 (preview) | Not stated by Google | Not stated by Google | Not stated by Google | No | Not stated by Google | Not stated by Google |
Release dates are Google Cloud GA dates; Gemini 3.1 Pro shows its preview launch.
Features stated across current models
- Thinking levels: Hydrogen supports low, medium, high and xhigh; Gemini 3.8 Flash supports low, medium and high. Carbon defaults to MEDIUM and its cyber variant to HIGH.
- Tools: function calling, structured outputs, code execution, computer use, file search, search and Maps grounding and URL context (Hydrogen, 3.8 Flash, 3.5 Flash-Lite pages).
- Caching: Gemini 4 Argon and Hydrogen support implicit caching only; explicit caching is not supported on those models.
- Consumption options: Batch API, Flex inference and Priority inference are listed for Hydrogen and 3.5 Flash-Lite.
Gemini model families and naming
- Gemini 4 (Argon, Carbon, Hydrogen): element-named models. Google's blog positions Carbon for the most demanding tasks and Argon as the recommended choice for most software engineering and knowledge work; Hydrogen is described as built for speed and efficiency.
- Flash and Flash-Lite: version-numbered models (3.1 to 3.8). Flash-Lite is positioned for low latency and cost.
- Pro: Gemini 3.1 Pro remains in preview; Gemini 3 Pro Preview was shut down on 9 March 2026 and now routes to gemini-3.1-pro-preview.
- Cyber variants: gemini-4-argon-cyber-permissive, gemini-4-carbon-cyber-permissive and gemini-3.8-flash-cyber are versions for defensive cybersecurity use cases such as vulnerability discovery and patch generation.
- Specialised models: Live, TTS, Transcribe, Nano Banana (image), Embedding, Robotics ER and Omni Flash (video).
- Preview and alias IDs: preview models carry -preview in the ID. The alias gemini-flash-latest pointed to gemini-3.5-flash from 19 May 2026, per the changelog. See LLM Naming and Versioning Explained.
When was each Gemini model released?
As of 2026-11-15. Dates are from the Gemini API changelog and the Google Cloud model versions page, in date order.
| Date | Event | API ID |
|---|---|---|
| 2025-06-17 | Gemini 2.5 Pro and 2.5 Flash released (Google Cloud) | gemini-2.5-pro, gemini-2.5-flash |
| 2026-02-19 | Gemini 3.1 Pro preview launched | gemini-3.1-pro-preview, gemini-3.1-pro-preview-customtools |
| 2026-05-07 | Gemini 3.1 Flash-Lite GA | gemini-3.1-flash-lite |
| 2026-05-19 | Gemini 3.5 Flash GA | gemini-3.5-flash |
| 2026-07-21 | Gemini 3.6 Flash and 3.5 Flash-Lite GA | gemini-3.6-flash, gemini-3.5-flash-lite |
| 2026-08-13 | Gemini 3.7 Flash GA | gemini-3.7-flash |
| 2026-08-27 | Gemini Omni Flash GA | gemini-omni-1.1-flash |
| 2026-09-02 | Gemini 3.8 Flash GA | gemini-3.8-flash |
| 2026-09-18 | Gemini 2.5 models limited to past users on the Gemini API | gemini-2.5-pro, -flash, -flash-lite |
| 2026-10-08 | Gemini 3.7 Flash and 3.5 Flash deprecated on the Gemini API, requests rerouted | gemini-3.7-flash, gemini-3.5-flash |
| 2026-10-13 | Gemini 4 Argon GA; Argon cyber permissive released | gemini-4-argon, gemini-4-argon-cyber-permissive |
| 2026-11-05 | Gemini Omni 1.2 Flash released (Google Cloud) | gemini-omni-1.2-flash |
| 2026-11-10 | Gemini 4 Carbon GA (Google Cloud); blog announcement 2026-11-11 | gemini-4-carbon, gemini-4-carbon-cyber-permissive |
| 2026-11-12 | Gemini 4 Hydrogen GA (Google Cloud) | gemini-4-hydrogen |
The latest release is Gemini 4 Hydrogen, on 12 November 2026.
How much do Gemini models cost?
As of 2026-11-15. USD per million tokens, Gemini API paid tier, standard processing. Batch shows input / output. Arrows mark a scheduled price change on the date shown. Source: Google AI Studio pricing page and the Gemini 4 Carbon announcement.
| Model | Input | Cached input | Output | Batch | Free tier | Notes |
|---|---|---|---|---|---|---|
| Gemini 4 Hydrogen | Not stated by Google | Not stated by Google | Not stated by Google | Not stated by Google | Not stated by Google | Batch, Flex and Priority listed as options |
| Gemini 4 Carbon | $2 → $4 (1 Dec 2026) | Not stated by Google | $10 → $20 (1 Dec 2026) | Not stated by Google | Paid plan required | Introductory price matches Argon until 30 Nov 2026 (blog) |
| Gemini 4 Argon | $2 → $4 (1 Dec 2026) | $0.10 → $0.20 | $10 → $20 | $1 / $5 → $2 / $10 | No | Implicit caching only; Priority $3.60 / $18 → $7.20 / $36 |
| Gemini Omni Flash | $1.50 | Not listed | $9 text; $17.50 video | Not listed | No | Video generation and editing |
| Gemini 3.8 Flash | $0.75 → $1.50 (1 Jan 2027) | $0.075 → $0.15 | $3.75 → $7.50 | $0.375 / $1.875 → $0.75 / $3.75 | Yes | Cache storage $0.50 → $1.00 per million tokens per hour |
| Gemini 3.6 Flash | $0.75 → $1.50 (1 Jan 2027) | $0.075 → $0.15 | $3.75 → $7.50 | $0.375 / $1.875 → $0.75 / $3.75 | Yes | Same rates as 3.8 Flash |
| Gemini 3.5 Flash-Lite | $0.30 | $0.03 | $2.50 | $0.15 / $1.25 | Yes | Cache storage $1.00 per million tokens per hour |
| Gemini 3.1 Flash-Lite | $0.25 ($0.50 audio) | $0.025 ($0.05 audio) | $1.50 | $0.125 / $0.75 | Yes | Audio input priced separately |
| Gemini 3.1 Pro (preview) | $2 (up to 200K); $4 (over 200K) | $0.20 / $0.40 | $12 / $18 | $1 / $6 (up to 200K); $2 / $9 | No | Tiered by prompt length at 200K |
| Gemini 3 Flash (preview) | $0.50 ($1.00 audio) | $0.05 ($0.10 audio) | $3.00 | $0.25 / $1.50 | Yes | Audio input priced separately |
| Gemini 2.5 Pro | $1.25 (up to 200K); $2.50 | $0.125 / $0.25 | $10 / $15 | $0.625 / $5 (up to 200K) | Yes | Past users only |
| Gemini 2.5 Flash | $0.30 ($1.00 audio) | $0.03 | $2.50 | $0.15 / $1.25 | Yes | Past users only |
| Gemini 2.5 Flash-Lite | $0.10 ($0.30 audio) | $0.01 | $0.40 | $0.05 / $0.20 | Yes | Past users only |
- Flex and Priority tiers: on Gemini 4 Argon, Flex matches Batch prices and Priority is $3.60 input and $18 output, rising to $7.20 and $36 from 1 December 2026.
- Grounding: Gemini 3 and later models share 5,000 free Google Search grounding requests per month, then $14 per 1,000; Gemini 2.5 models get 1,500 free requests per day, then $35 per 1,000 grounded prompts.
See Model Pricing Explained and How to set spend limits on LLM API keys.
Where are Gemini models available?
- Gemini API and Google AI Studio: developer access with free and paid tiers; Gemini 4 Argon and Carbon require the paid tier.
- Gemini Enterprise Agent Platform (Google Cloud): Gemini 4 Argon in global, us, eu, europe-west2 and asia-southeast1; Gemini 4 Carbon and Hydrogen in global, us and eu.
- Consumer apps: Google AI Ultra subscribers receive Gemini 4 Carbon in the Gemini app, Google Antigravity and AI Mode in Google Search, per the Carbon announcement.
- Business: the Carbon announcement lists Gemini for business and CodeMender as enterprise routes, and Android Studio for paid-tier developers.
See the Google Gemini API and AI Studio Security Guide and the Google Vertex AI Security Guide.
Restricted-access Gemini models
As of 2026-11-15. Source: Google Cloud model pages for Gemini 4 Argon and Carbon, and the Carbon announcement.
| Model | Access stated by provider | Data retention condition |
|---|---|---|
| Gemini 4 Carbon (Google Cloud) | Verified enterprise customers in select regions; API on the paid tier; consumer access with Google AI Ultra | Not stated by Google |
| Gemini 4 Carbon for cyber (gemini-4-carbon-cyber-permissive) | Allowlist only through the Fairwind Program; global and us regions; computer use not supported | Not stated by Google |
| Gemini 4 Argon Cyber Permissive (gemini-4-argon-cyber-permissive) | Allowlist only through the Fairwind Program; a Google Frontier Cyber Model subject to Section 31 of the Service Specific Terms | Not stated by Google |
| Gemini 3.8 Flash Cyber (gemini-3.8-flash-cyber) | Separate application; access does not carry over to the Gemini 4 cyber variants | Not stated by Google |
| Gemini 2.5 Pro, Flash, Flash-Lite (Gemini API) | Limited to users who used them previously, since 18 September 2026 | Not stated by Google |
What is the context window of Gemini models?
Gemini 4 Argon, Carbon and Hydrogen each have a 1,048,576-token context window shared by input and output. A single foreground segment returns up to 65,536 tokens when streaming or 32,768 when not streaming on Argon and Carbon, and Hydrogen's default output limit is 65,536 tokens; long decode extends output up to the full window. Gemini 3.8 Flash and 3.5 Flash-Lite list 1,048,576 input tokens and 65,536 output tokens. See Context Windows Explained.
Is any Gemini model open source?
No Gemini model is published with open weights. Google's open-weight family is Gemma: Gemma 4 comes in E2B, E4B, 12B, 26B A4B (a mixture-of-experts model activating about 4B parameters per token) and 31B sizes, with downloads on Kaggle and Hugging Face. Google's Gemma documentation says the weights permit responsible commercial use under the Gemma 4 licence, small models support 128K context and medium models 256K. Gemma 4 is free on the Gemini API. See Google Gemma Models and Running LLMs Locally Explained.
Model cards and system cards
As of 2026-11-15.
| Model | Model card or page | System card | Technical report |
|---|---|---|---|
| Gemini 4 Argon, Carbon, Hydrogen | Google Cloud model pages and AI Studio model pages (ID, context, output, modalities, regions) | Not stated by Google | Not stated by Google |
| Gemini 3.8 Flash, 3.5 Flash-Lite | AI Studio model pages | Not stated by Google | Not stated by Google |
| Other Gemini models | AI Studio models list entry | Not stated by Google | Not stated by Google |
The Google Cloud model pages act as Google's specification pages for each version. See Model Cards and System Cards Explained.
Benchmarks
As of 2026-11-15. Provider-reported and independent results are kept in separate columns.
| Model | Provider-reported benchmarks | Independent leaderboard results |
|---|---|---|
| Gemini 4 Carbon | Company-reported in the announcement (2026-11-11): first place on CursorBench 4.0, WANDR and AutomationBench; 92% on the 375-task Agentic Honesty Benchmark (Argon 70%) | None cited on this page |
| Gemini 4 Argon | Not stated on the pages checked | None cited on this page |
| Gemini 4 Hydrogen | Not stated on the pages checked | None cited on this page |
Deprecations and retirements
As of 2026-11-15. Source: Google Cloud model versions page and Gemini API changelog. Google Cloud retirement dates for GA models are at least 12 months after release and may be extended but not brought forward.
| Model | API ID | Deprecated or notice | Retirement or shutdown | Stated replacement |
|---|---|---|---|---|
| Gemini 3.6 Flash | gemini-3.6-flash | Shorter availability model (Google Cloud) | 2026-11-19 (Google Cloud); Stable on Gemini API | gemini-3.8-flash |
| Gemini 2.5 Flash Live | gemini-live-2.5-flash-native-audio | Listed on Google Cloud | 2026-12-13 (Google Cloud) | gemini-3.8-live |
| Gemini 3.7 Flash | gemini-3.7-flash | 2026-10-08 (Gemini API, rerouted to gemini-3.8-flash) | 2027-01-28 (Google Cloud) | gemini-3.8-flash |
| Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-08 (Gemini API, rerouted to gemini-3.6-flash) | 2027-05-19 or later (Google Cloud) | Not stated by Google |
| Gemini 2.5 Pro | gemini-2.5-pro | Past users only on Gemini API since 2026-09-18 | 2026-10-20 (Google Cloud, retired) | gemini-3.8-flash or gemini-3.5-flash |
| Gemini 2.5 Flash | gemini-2.5-flash | Past users only on Gemini API since 2026-09-18 | 2026-10-20 (Google Cloud, retired) | gemini-3.8-flash or gemini-3.5-flash-lite |
| Gemini 2.5 Flash-Lite | gemini-2.5-flash-lite | Past users only on Gemini API since 2026-09-18 | 2026-10-20 (Google Cloud, retired) | gemini-3.8-flash, gemini-3.5-flash-lite or Gemma 4 |
| Gemini 2.0 Flash and Flash-Lite | gemini-2.0-flash, gemini-2.0-flash-lite | 2026-02-18 | 2026-06-01 (shut down) | gemini-3.5-flash-lite |
| Gemini 3 Pro Preview | gemini-3-pro-preview | 2026-02-26 | 2026-03-09 (shut down) | gemini-3.1-pro-preview |
The AI Studio models page states that the Gemini 2.5 models are not deprecated on the Gemini API and remain available until further notice to previous users. Gemini 4 Argon carries a Google Cloud retirement date of 13 October 2027 or later; Carbon and Hydrogen have no retirement date announced. See the AI Model Retirement Guide.
Data handling and security
- Unpaid services: the Gemini API Additional Terms state that Google uses content submitted to unpaid services and generated responses to provide, improve and develop products, that human reviewers may read and annotate that data after it is disconnected from the account, and that users should not submit sensitive, confidential or personal information.
- Paid services: Google states it does not use prompts or responses to improve its products; prompts and responses are processed under the Data Processing Addendum and logged for a limited period solely to detect Prohibited Use Policy violations and for required legal disclosures.
- EEA, Switzerland and UK: the paid-service data terms apply to all services, including AI Studio and unpaid API quota, and only paid services may be used when making API clients available to users in those regions.
- Grounding: for Grounding with Google Search, Google stores prompts, contextual information and output for 30 days.
- Stored requests: from 26 October 2026, new paid projects using the Generate Content API default to storing requests (store=true), per the changelog; existing projects keep their settings.
See AI Data Retention and Deletion and the Gemini Enterprise Security Guide.
How to choose a Gemini model: factual criteria
- Access: Carbon on Google Cloud needs enterprise verification; cyber variants need Fairwind allowlisting; Argon, Hydrogen and the Flash models are generally available.
- Price timing: budget for the 1 December 2026 change on Argon and Carbon and the 1 January 2027 change on 3.8 and 3.6 Flash.
- Data terms: free tier content may be used to improve products; paid tier content is not, per the terms.
- Context and output: all Gemini 4 models offer 1,048,576 tokens with long decode; Flash and Flash-Lite list 65,536 output tokens.
- Modalities: real-time audio, speech output, images and video use the separate Live, TTS, Nano Banana and Omni models.
- Lifecycle: check the Google Cloud retirement date and the Gemini API status separately before building on a version.
- Region: confirm the model is offered in the required Google Cloud region; Argon lists more regions than Carbon and Hydrogen.
Practitioner checklist
- Pin production code to an explicit Gemini model ID rather than a -latest alias and record it in the AI inventory.
- Move workloads off gemini-3.6-flash on Google Cloud before 19 November 2026.
- Re-estimate Gemini 4 Argon and Carbon costs for the 1 December 2026 price change.
- Confirm that rerouted IDs (gemini-3.7-flash, gemini-3.5-flash) still meet quality and cost expectations.
- Use the paid tier for any sensitive data and for any application serving users in the EEA, Switzerland or the UK.
- Review the store=true default on new paid projects against the organisation's retention policy.
- Store Gemini API keys in a secrets manager and set spend limits. See the LLMjacking Guide.
Standards and references
- Google: Gemini API models (accessed 2026-11-15)
- Google: Gemini API pricing (accessed 2026-11-15)
- Google Cloud: Model versions and lifecycle (accessed 2026-11-15)
- Google: Gemini API release notes (accessed 2026-11-15)
- Google Cloud: Gemini 4 Carbon (accessed 2026-11-15)
- Google Cloud: Gemini 4 Argon (accessed 2026-11-15)
- Google Cloud: Gemini 4 Hydrogen (accessed 2026-11-15)
- Google: Gemini 4 Hydrogen model page (accessed 2026-11-15)
- Google: Gemini 3.8 Flash model page (accessed 2026-11-15)
- Google: Gemini 3.5 Flash-Lite model page (accessed 2026-11-15)
- Google: Gemini 4 Carbon announcement (accessed 2026-11-15)
- Google: Gemini API Additional Terms of Service (accessed 2026-11-15)
- Google: Gemma 4 overview (accessed 2026-11-15)
Related NHI Mgmt Group resources: Gemini CLI and Gemini Code Assist Security Guide, Gemini for Google Workspace Security Guide, Reasoning Models Explained, Multimodal LLMs Explained.