AI Wiki · Mediocre Security
Model familyModel families

Gemini

Google DeepMind's multimodal frontier model family, launched December 2023; Gemini 4 Argon and the cyber-restricted Gemini 4 Carbon arrived in fall 2026.

Gemini is Google DeepMind’s family of natively multimodal models, the successor to PaLM and the brand that replaced the Bard chatbot in early 2024. In the Gemini app, Personal Intelligence connects Gmail, Photos, YouTube and Search data; Google Wallet was added for all US users on Nov 13 2026, with optional bank and card linking through Plaid, for budgeting and spending questions. From mid-November 2026 Google keeps chats made with Gemini History off for around two months instead of 72 hours, for responses, feedback and safety review including by human reviewers (Gizmodo, Nov 17 2026).

Version history

Model Released Notes
Bard (PaLM 2) Mar 2023 Google’s rushed response to ChatGPT
Gemini 1.0 (Ultra, Pro, Nano) Dec 2023 First natively multimodal Gemini
Gemini 1.5 Feb 2024 1M-token context window
Gemini 2.0 Dec 2024 “Agentic era” framing
Gemini 2.5 Mar 2025 Thinking models; 2.5 Flash later used as a fallback in the VOLTA infostealer
Gemini 3 Nov 2025
Gemini 3.7 Flash 2026 Current fast tier
Gemini 4 Argon Sep 30 2026 (GA Oct 13) 77.9% DeepSWE v1.1; 1M-token output ceiling; $2/$10 per M tokens intro
Gemini 4 Carbon Oct 28 2026 (GA Nov 10) Cyber model first restricted to vetted defenders (Fairwind Program); topped MathArena at 85.4% on Nov 14, at ~15x GPT-6.1 Sol’s cost
Gemini Omni 1.2 Flash Nov 5 2026 Video generation, GA on the Gemini API; 15-second clips, $0.07/s at 720p
Gemini Decision 1 Nov 11 2026 Scoring model behind the Gemini Decisions API: ranks options in one pass; $0.10 per M input tokens, no output charge
Gemini 4 Neon Nov 18 2026 Mid tier between Carbon and Hydrogen; $3/$15 per M tokens, 1M context; 61 on the Artificial Analysis Intelligence Index, 90.5% SEC-bench Pro (Google’s figures); cyber variant via Fairwind
Gemini 4 Hydrogen Nov 12 2026 $1.50/$7.50 per M tokens, 1M-token context; 51 on the Artificial Analysis Intelligence Index (Google’s figure)

Restricted releases

Google’s Fairwind Program gives vetted cyber defenders first access to its most capable models, including a version of Argon without cyber guardrails and the restricted Gemini 4 Carbon (88.2% SEC-bench Pro, 71.0% CWE-bench), while the US government runs voluntary pre-release testing. It is the clearest example of defender-first staged release, an approach OpenAI mirrors with Daybreak. See also the Gemini CTF incident.

Agents

On Nov 19 2026 Google added computer use to the Gemini Mac app: it operates native apps and browses in the background in a live preview where the user can pause, stop or approve actions. It launched English-only in the US for Google AI Plus, Pro and Ultra subscribers (TechCrunch).

Timeline

Every dated event in the wiki that involves Gemini, newest first.

  • Release
    Google DeepMind launches Gemini 4 Neon, the mid tier of Gemini 4

    Neon sits between Carbon and Hydrogen at $3 per million input tokens ($0.15 cached) and $15 output, with a 1M-token context. Google claims 61 on the Artificial Analysis Intelligence Index (third overall), 70.9% on the Vals Index, 65.2% on Terminal-Bench 4.1 and 90.5% on SEC-bench Pro (vendor claims). It is on the Gemini API, Gemini for business, Antigravity and the Gemini app for Pro and Ultra subscribers; a cyber-defense variant goes to vetted defenders through the Fairwind Program. The same day Google opened Genie 4, its interactive world model, to AI Ultra subscribers in Project Genie.

  • Release
    Reflection AI gates its Laser autonomous pentest agent behind a Trusted Access Program

    Reflection's program gives vetted organizations early access to selected capabilities for defensive cybersecurity, starting with Laser, an autonomous penetration-testing harness for its Mirror terminal agent offered as a research preview only to accepted commercial partners. Laser scopes tests, keeps a shared evidence ledger and uses a separate supervisor to review evidence and direct the executor. It follows gated cyber releases such as OpenAI's Daybreak and Google's Fairwind Program for Gemini.

  • Policy
    Google now keeps Gemini chats about two months even with history turned off

    Google's updated Gemini Apps Privacy Hub raises retention for chats made with Gemini History off from 72 hours to around two months, used to respond, process feedback and for safety, with help from human reviewers. Gizmodo compares OpenAI (up to 30 days for temporary chats) and Anthropic (30 days by default). Google declined to comment beyond a blog post on updated Gemini controls.

  • Research
    Study: ChatGPT, Gemini and Perplexity point shoppers to risky retail sites

    Brand-protection firm Podqi asked the three chatbots shopping questions about 151 brands across 6,795 answers in September and found fake, re-registered or dead store domains in 15% of Perplexity answers, 8% of ChatGPT answers and 4% of Gemini answers (counting mentions that warned about the site). Podqi wrote its own prompts, and one of the brands is its client; results are one firm's test, shared with The Verge.

  • Release

    A cheaper Gemini 4 tier at $1.50 / $7.50 per million tokens with a 1M-token context. Google cites 51 on the Artificial Analysis Intelligence Index (Argon: 50) and 2.3x more resilience on Gray Swan's indirect Prompt injection benchmark (vendor claims).

    Sources: Google, [link]
  • Release
    Google makes Gemini 4 Carbon generally available

    Carbon, first released to cyber defenders in the Fairwind Program two weeks earlier, opens to paying developers and enterprises via the API and rolls out to AI Ultra subscribers. Introductory price $2 / $10 per million tokens until Nov 30, then $4 / $20. Google claims top ranks on CursorBench 4.0 and several agent and cyber benchmarks (vendor claims).

  • Release
    Gemini 4 Argon generally available

    Gemini app for AI Pro/Ultra; API from $2/$10 per million tokens, doubling after Nov 30.

    Sources: Google; Reuters

Connections

Gemini and everything one link away. Drag to move, scroll to zoom, click a node to open it.Open in the full graph →

What links here

Legal action AI copyright litigation (2026)

Company Apple · Google · OpenRouter

Product ChatGPT · DeepSeek Harness (dsh) · IBM watsonx Orchestrate

Government body European Union

Incident Gemini CTF incident (Irregular) · VOLTA infostealer ranks victims with LLMs

Model family Gemma

AI lab Google DeepMind · Reflection AI

Protocol Google Universal Cart (Universal Commerce Protocol) · Model Context Protocol (MCP)

Concept Large language model (LLM) · Multimodal · Non-human identity (NHI)

Sources

  1. Wikipedia, “Gemini (language model)”
  2. Google [link], Nov 12 2026
  3. Google, “Gemini Decisions API: Fast, calibrated multimodal decisions”, Nov 11 2026
  4. Google, “Connect Google Wallet to Gemini for personal finance insights”, Nov 13 2026
  5. The Decoder, “Gemini 4 Carbon takes the MathArena crown but costs 15 times more than GPT-6.1 Sol”, Nov 14 2026
  6. Google, “Introducing Gemini 4 Neon”, Nov 18 2026
  7. Gizmodo, “Google to Keep AI Chat Logs for Two Months, Even With Gemini History Turned Off”, Nov 17 2026
  8. TechCrunch, “Google gives Gemini the ability to use your Mac for you”, Nov 19 2026