Google Gemini

Google Gemini is the multimodal model family from Google DeepMind. It processes text, images, audio, video and PDF and is available through the Gemini API (AI Studio) and the Gemini Enterprise Agent Platform (formerly Vertex AI). For EU enterprises the Agent Platform is what matters, because data residency and contractual commitments are documented only there. The relevant models today are Gemini 4 Argon, GA since 13 October 2026 and the first model of the Gemini 4 generation, and the Flash line with Gemini 3.8 Flash and 3.5 Flash-Lite; Gemini 3.6 Flash has a retirement date of 19 November 2026 on the Agent Platform. Gemini 3.7 Flash and 3.5 Flash have been deprecated in the Gemini API since 8 October 2026 but keep running on the Agent Platform. Gemini 3.1 Pro remains listed as a preview; Gemini 3.5 Pro has not shipped.

innFactory AI Consulting, based in Rosenheim, Germany, supports enterprises across the DACH region and Europe in selecting, integrating and using Google Gemini in a GDPR-compliant way, including through CompanyGPT. This overview reflects the state as of October 2026.

Current models at a glance

Prices per Google (Gemini API, global endpoint); on the Agent Platform, prices outside global are 10% higher (see pricing). “Gemini API: deprecated” means no longer selectable in AI Studio, with requests automatically routed to 3.8 Flash or 3.6 Flash respectively; on the Gemini Enterprise Agent Platform both models remain usable until their retirement (see lifecycle). Gemini 3.6 Flash has a retirement date of 19 November 2026 on the Agent Platform (replacement: Gemini 3.8 Flash); the Gemini API names no shutdown date.

Models in detail

Gemini 4 Argon – frontier model of the Gemini 4 generation

Gemini 4 Argon (gemini-4-argon) has been GA since 13 October 2026: in the Gemini API, AI Studio, on the Gemini Enterprise Agent Platform, in Antigravity and, for Google AI Pro and Ultra subscribers, in the Gemini app and AI Mode in Search. Google had announced the model on 30 September 2026, initially only for vetted cyber defenders in the Fairwind Program. Google positions Argon as a frontier model for “deep work” in coding, financial and legal analysis, cybersecurity and science. According to the model card, it is a newly trained sparse mixture-of-experts model, not a fine-tune of a predecessor; the knowledge cutoff is March 2026.

The figures published by Google (model evaluation PDF, pass@1, highest thinking setting) compared with Gemini 3.8 Flash:

The 3.8 Flash figures come from Google’s September 2026 evaluation. In the Argon PDF, Google also lists figures for GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5, mostly taken from their self-reported numbers or public leaderboards. On FrontierSWE v2, Terminal-bench 4.0, PostTrainBench, Terminal-Bench Science and OSWorld 2.0, other models lead according to the same table; on CWE-bench v1 the top scores are tied. Independent measurements were not yet available at GA.

Technical specs: input, thinking and output share one 1,048,576-token context window. A single request generates up to 65,536 tokens (32,768 for non-streaming direct HTTP calls); long decodes via continuation_token allow up to 1,048,576 output tokens, per Google currently only through API version v1beta1. The model accepts text, images, video, audio and PDF and outputs text only. Reasoning is controlled with the thinking levels low, medium (default) and high. Available tools include function calling, structured output, code execution, URL context, grounding with Google Search and Google Maps, file search, agentic video understanding and computer use (preview), plus batch, flex, priority and Provisioned Throughput. Tuning, the Live API and explicit context caching are not supported.

Pricing (Gemini API): until 30 November 2026, Argon costs $2 per 1M input tokens and $10 per 1M output tokens (including thinking tokens; caching $0.10; batch and flex $1 / $5; priority $3.60 / $18). From 1 December 2026 prices rise to $4 / $20 (caching $0.20). There is no context-length tier and no free tier. Per Google, every processed token is billed, including internal tool system instructions and function declarations when grounding. On the Gemini Enterprise Agent Platform, Argon costs the same in the global endpoint; outside global, including the EU multi-region endpoint, the price is $2.20 / $11 (from 1 December 2026 $4.40 / $22). There, Google implements the promotional price as a 50% credit back on net spend. From 13 November 2026, Google bills the retrievable tokens of completions longer than 32K tokens on the Agent Platform even if the request ultimately fails due to a client or server timeout.

Migration note: Google publishes a guide for moving from Gemini 3.1 Pro. Key points: thinking_budget and thinking_level: MINIMAL return HTTP 400; temperature, top_p and top_k are ignored; candidate_count, frequency_penalty, presence_penalty and prefilled model turns must be removed. Explicit caching goes away; static prompt content of 6,144 tokens or more belongs at the start so implicit caching applies. A current Google Gen AI SDK is required (Python google-genai 2.30.1 or later).

Cyber variant: Gemini 4 Argon Cyber Permissive (gemini-4-argon-cyber-permissive) is meant for defensive cybersecurity such as vulnerability discovery, reverse engineering, exploit verification and patch generation. It is GA but reachable only via allowlist through the Fairwind Program, runs only in global and us, and as a “Google Frontier Model” is subject to prompt and response logging for abuse monitoring (up to 30 days inside the customer environment). Per Google’s documentation, this special logging does not apply to the standard model gemini-4-argon.

Gemini 3.8 Flash – current Flash model

Gemini 3.8 Flash (gemini-3.8-flash) has been available straight to GA since 2 September 2026 – with no preview and no preview suffix in the model ID. Google positions it as its most intelligent Flash model to date and stresses that it “works harder” on complex tasks: it runs additional reasoning steps and calls tools iteratively. For cost planning that is the most important framing, because the list price is identical to 3.7 Flash while token consumption per task is higher (see below).

Google’s published benchmark figures (model evaluation PDF, pass@1) compared with Gemini 3.7 Flash:

The DeepSWE figure is confirmed by the independent DeepSWE leaderboard (74% ± 1). In the Artificial Analysis Intelligence Index 3.8 Flash scores 59 points (3.7 Flash: 56, 3.6 Flash: 52). The same measurement also shows that 3.8 Flash generates around 30% more output tokens per task than 3.7 Flash, raising cost per task by roughly 40% at the same list price. Anyone migrating from 3.7 to 3.8 Flash should therefore measure cost per task rather than compare token prices.

The context window is 1,048,576 input tokens and 65,536 output tokens; the knowledge cutoff per the model card is March 2026. The model processes text, image, video, audio and PDF; output is text only (no Live API). Reasoning is controlled via the thinking levels low, medium (default) and high; minimal is rejected with an error, and thinking_budget has been replaced by thinking_level. Tooling covers function calling, structured output, code execution, file search, URL context, Google Search grounding, batch, flex and priority inference, and computer use (preview). The agentic video understanding introduced on 1 September 2026 is explicitly listed by Google only for 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite – not for 3.8 Flash.

Pricing (Gemini API): Through 31 December 2026 the same introductory price as for 3.7 and 3.6 Flash applies: $0.75 per 1M input tokens and $3.75 per 1M output tokens (including thinking tokens; caching $0.075; batch $0.375 / $1.875; priority $1.35 / $6.75). From 1 January 2027 list prices rise to $1.50 / $7.50 / $0.15. Per Google the introductory price applies both in AI Studio and on the Gemini Enterprise Agent Platform. There is no context-length tiering; audio input is not priced separately.

Migration note: Besides temperature, top_p and top_k, candidate_count is deprecated as well; FunctionResponse objects require call_id and name. Existing integrations on 3.5 or 3.6 Flash that still send thinking_level: minimal break on the switch.

Gemini 3.7 Flash

Gemini 3.7 Flash (gemini-3.7-flash) has been generally available since 13 August 2026 – shipped straight to GA with no preview phase. Google positions the model for coding, web development, knowledge work, and agentic workflows, and emphasises that the gains come not from a retraining run but from algorithmic improvements and evaluated developer feedback.

Google’s published benchmark results compared to Gemini 3.6 Flash:

The context window is 1,048,576 input tokens and 65,536 output tokens. The model processes text, image, video, audio, and PDF; reasoning is controlled via the thinking levels low, medium, and high (there is no “minimal” level). Tooling covers function calling, structured output, code execution, file search, Google Search grounding, and computer use (preview). Relevant for media workloads: since 1 September 2026 agentic video understanding is GA for 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite – at no surcharge, per Google with up to 88% fewer tokens, up to 66% lower cost and up to 7% better accuracy on LongVideoBench.

Pricing (Gemini API): Through 31 December 2026 an introductory price of $0.75 per 1M input tokens and $3.75 per 1M output tokens applies (caching $0.075). From 1 January 2027 list prices rise to $1.50 / $7.50 / $0.15 – anyone building a business case now should factor that doubling in. A limited free tier is available.

The knowledge cutoff per the model card is March 2026. In the Gemini API, 3.7 Flash has been deprecated since 8 October 2026: the model can no longer be selected in AI Studio, and requests to gemini-3.7-flash are automatically served by gemini-3.8-flash. On the Gemini Enterprise Agent Platform, 3.7 Flash runs until its retirement on 28 January 2027. Migration note: the sampling parameters temperature, top_p, and top_k have been deprecated since 21 July 2026 and this applies to 3.7 Flash as well – review existing integrations.

Gemini 3.6 Flash

Gemini 3.6 Flash (gemini-3.6-flash) improves coding, knowledge work, and multimodal processing over 3.5 Flash. On the DeepSWE coding benchmark the model jumps from 37 to 49 percent, and it reaches 83 percent on OSWorld-Verified (computer use). At the same time it produces 17 percent fewer output tokens than 3.5 Flash – Google directly addressing developer feedback on verbosity, which lowers effective per-task cost further. The knowledge cutoff is March 2026; the context window is 1 million input tokens (65,536 output).

The introductory price is $0.75/1M input and $3.75/1M output, limited to 31 December 2026 (launch price: $1.50/$7.50). The model is available via the Gemini API (AI Studio, Android Studio), Antigravity, the Gemini Enterprise Agent Platform and the Gemini app. On the Agent Platform the model page lists a retirement date of 19 November 2026, with gemini-3.8-flash as the replacement; production workloads on 3.6 Flash should be moved in good time.

Gemini 3.5 Flash-Lite – for high-volume workloads

Gemini 3.5 Flash-Lite (gemini-3.5-flash-lite) is the fastest model in the 3.5 series: 350 output tokens per second as measured by Artificial Analysis, priced at $0.30/1M input and $2.50/1M output. That makes it a fit for high-volume automation, classification, and latency-critical applications – for example as a sub-agent in multi-agent pipelines. Google is additionally rolling the model out in Google Search.

Gemini 3.8 Flash Cyber and the Fairwind Program – security models with restricted access

With Gemini 3.5 Flash Cyber, Google had announced its first dedicated cybersecurity model on 21 July 2026, paired with the CodeMender security agent and exclusive to governments and vetted partners. On 2 September 2026 the successor Gemini 3.8 Flash Cyber followed, provided exclusively through the new Fairwind Program – for governments, critical infrastructure operators and core technology platforms, per Google with over 650 partners at launch. Google reports for 3.8 Flash Cyber, among other figures, 86.2% pass@1 on CyberGym (3.5 Flash Cyber: 77.5%), 71.0% on an internal vulnerability benchmark across 20 programming languages (3.7 Flash: 58.9%) and 47.2% pass@1 on CWE-Bench. Since 13 October 2026 the same programme also provides Gemini 4 Argon Cyber Permissive (see above); per the Argon blog, Google is expanding Fairwind to hundreds more partners. For enterprises outside the programme the cyber models are not available, but they show Google’s course of rolling out security-critical capabilities in a controlled way – a pattern we already know from the clearance processes at OpenAI GPT-5.6 and Anthropic Claude Mythos 5.1.

Gemini 3.5 Flash – agentic-first, with an EU single region

With Gemini 3.5 Flash, Google positioned a Flash model above its own Pro model for the first time in May 2026: 3.5 Flash outperforms Gemini 3.1 Pro on Terminal-Bench 2.1 (coding), GDPval-AA Elo (real-world agentic) and MCP Atlas (tool use). The model is built for multi-hour autonomous workflows and pauses for human input at decision points.

For EU projects, 3.5 Flash stays relevant because it is the only current Gemini model with a single-region endpoint in Frankfurt (europe-west3) and London (europe-west2, UK) in addition to the EU multi-region. In the Gemini API it has been deprecated since 8 October 2026, with requests to gemini-3.5-flash automatically routed to gemini-3.6-flash. On the Gemini Enterprise Agent Platform, 3.5 Flash remains GA; its model page gives the retirement as “May 19, 2027 or later”.

Gemini 3.1 Pro and Gemini 3.5 Pro

Gemini 3.1 Pro (gemini-3.1-pro-preview, preview since 19 February 2026, $2/$12 per 1M tokens up to 200k context, up to 2M context) was the family’s Pro model until the Argon launch. It remains listed without a shutdown date but is usable only through the global endpoint, with no documented EU data residency. For the switch, Google points to its migration guide for Gemini 4 Argon.

Gemini 3.5 Pro was announced at Google I/O on 19 May 2026 as an orchestrator/planner for multi-agent workflows with a 2M token context and a Deep Think mode, but has not shipped and has no date. There is no Gemini 3.6, 3.7 or 3.8 Pro; those generations are Flash-only lines.

Gemini Spark – agentic 24/7 assistant

At Google I/O 2026, Google introduced Gemini Spark, an agentic personal assistant; it is now built on Gemini 3.7 Flash. Spark runs continuously on Google’s cloud, can be addressed directly via a dedicated Gmail address, and executes background tasks across Chrome and Google Workspace. Spark supports the Model Context Protocol (MCP) for external integrations.

Relevant for customers in Europe: Google states Spark availability in over 160 countries, but explicitly excludes the EEA, Switzerland, the United Kingdom, and Nigeria. That applies to the end-user product, not the API – Gemini 3.7 Flash itself is usable from the EU via the Gemini API and the Gemini Enterprise Agent Platform.

Very Large Context Window

Gemini 3.1 Pro processes up to 2 million tokens in a single context. This enables analysis of extensive contract portfolios, technical documentation, or entire codebases in a single request. The Flash models offer a 1,048,576 token context window at significantly lower cost.

Native Multimodal Processing

The Gemini 3.x family natively processes text, images, audio, video, and PDF documents. This enables use cases such as automated video analysis, document extraction from scanned PDFs, or meeting evaluation combining audio and visual material.

Google Ecosystem

• Gemini Enterprise Agent Platform (formerly Vertex AI): Enterprise deployment with SLA and data residency options
• Antigravity: Agentic development platform for the current Flash models
• Gemini Enterprise: Enterprise frontend (web, desktop, mobile) with access to Gemini 4 Argon and the current Flash models
• Google Workspace: Integration with Docs, Sheets, Gmail, and more
• Search Grounding: Access to current web information
• Google AI Studio: Rapid prototyping and API access

Pricing

List prices per 1M tokens (input/output) for the models currently relevant to enterprises:

The table shows prices in the global endpoint. Since 1 July 2026, Google charges 10% more on the Agent Platform for GA models from Gemini 3 onwards outside global, including the EU multi-region endpoint and single regions: for example $0.825 / $4.125 for Gemini 3.8 Flash through 31 Dec 2026, $1.65 / $9.90 for Gemini 3.5 Flash and $2.20 / $11 for Gemini 4 Argon through 30 Nov 2026 (pricing page). Tuned models cost 1.5 times the base model outside global.

For the Gemini 2.5 line (Pro, Flash, Flash-Lite) the Agent Platform lists 20 October 2026 as the retirement date, which has now passed; in the Gemini API it continues as a legacy generation without a shutdown date and with access limited to existing users.

Other models in the Gemini family

• Realtime and speech: Gemini 3.8 Live (gemini-3.8-live) and Gemini 3.8 Live Extended Thinking (gemini-3.8-live-extended-thinking) have been GA since 15 September 2026 (audio-to-audio via the Live API). Gemini 3.8 Flash TTS and 3.8 Flash-Lite TTS have been GA since 22 September 2026. Gemini 3.5 Transcribe and Transcribe Live (diarisation, word timestamps) have been available since 26 August 2026. Model pages exist on the Agent Platform for Gemini 3.8 Live, the 3.8 TTS models, 3.5 Transcribe and 3.5 Live Translate; we have not checked regions and residency for each. The predecessor previews gemini-3.1-flash-tts-preview, gemini-3.1-flash-live-preview, gemini-2.5-flash-native-audio-preview-12-2025, gemini-2.5-flash-preview-tts and gemini-2.5-pro-preview-tts are scheduled for shutdown on 17 November 2026.
• Video: Gemini Omni 1.1 Flash (gemini-omni-1.1-flash, GA since 27 August 2026) with scene extension up to 40 seconds, first/last-frame interpolation and 1080p/4K upscaling. On the Agent Platform, gemini-omni-1.1-flash-preview (preview, region global) is listed. For details and Veo availability see the Google Veo page.
• Image: Gemini Nano Banana 2.1 (gemini-nano-banana-2.1) has been GA in the Gemini API since 6 October 2026 and succeeds Nano Banana 2 (gemini-3.1-flash-image, deprecated, no shutdown date); also Nano Banana Pro (gemini-3-pro-image, up to 4K). On the Agent Platform, Nano Banana 2.1 is available in the global region only, per its model page. Details on the Nano Banana pages; Nano Banana (gemini-2.5-flash-image) is deprecated and listed only until 15 March 2027, see Gemini 2.5 Flash Image.
• Embeddings: Gemini Embedding 2 (gemini-embedding-2) has been stable since 22 April 2026 and maps text, images, video, audio and PDF into a shared vector space (8,192 input tokens, 128 to 3,072 dimensions, default 3,072, over 100 languages). Its predecessor gemini-embedding-001 (text only) shuts down on 14 May 2028. If you want to generate embeddings locally, the open EmbeddingGemma 2 is covered on the Google Gemma page.
• Agents and research: Antigravity Agent (antigravity-preview-09-2026), Deep Research and Deep Research Max (deep-research-preview-04-2026, deep-research-max-preview-04-2026) and Gemini Robotics ER 2 are in preview.

Data Residency: Keep DRZ and MLP Apart

The most important point for GDPR projects using Gemini is a distinction that Google’s documentation maintains consistently but that often blurs in marketing material:

• DRZ – data residency at rest: stored data stays in the chosen region.
• MLP – machine learning processing: the model’s processing also takes place in the chosen region.

Only both together amount to full EU localisation. The following status comes from the model pages of the Gemini Enterprise Agent Platform (checked on 28 October 2026, “Supported regions” section per model):

Region Model

On the Gemini Enterprise Agent Platform, Google distinguishes the global endpoint (default, best performance and newest features, but no residency guarantee) from the multi-region endpoints us and eu (https://aiplatform.eu.rep.googleapis.com for the EU) and from single regions such as europe-west3 (Frankfurt) or europe-west1 (Belgium). Which models run in which single region differs significantly per model and generation – Gemini 3.5 Flash is currently the only Flash model with an EU single-region endpoint, Gemini 4 Argon has single regions only in London (UK) and Singapore, while Gemini 2.5 Pro runs exclusively via single regions without its own eu multi-region endpoint.

Zero Data Retention

For the Gemini API / AI Studio, Google does not use prompts and responses from “paid services” to improve its products, per its ZDR page, but logs them for a limited period for abuse monitoring; in the free tier, data is used to improve products per the pricing page. For guaranteed zero data retention or enterprise data processing agreements, Google points to the Agent Platform. Not disableable is the 30-day retention for Google Search grounding and Google Maps grounding. Data is also stored by default in the Interactions API (store: true) and, for projects created on or after 26 October 2026, in the Generate Content API: new paid-tier projects store requests there as logs by default, retained for 55 days (configurable to 7, 14 or 28 days); switch this off with store: false or in the AI Studio log settings, while older projects keep their existing setting. On top of that come the Live API (session state for up to 24 hours with session resumption), explicit context caching (until the TTL expires) and the File API (until deletion or expiry).

On the Gemini Enterprise Agent Platform, ZDR is achievable per vendor documentation and practitioner reports, but requires active steps: data caching for Google models must be disabled and abuse monitoring logging must be turned off – the latter via opt-out through invoiced billing or as an exception via a request to Google support. For some “advanced AI features”, these sources indicate ZDR is not possible. We treat these statements as secondary-sourced and verify them contractually in each project.

The Gemini API Is Not the Enterprise Platform

The Gemini API / AI Studio is available in over 190 countries and therefore across the entire EU, but carries no EU data residency and no GDPR-specific commitments. Google explicitly points to the Gemini Enterprise Agent Platform for compliance commitments. For production processing of personal data, going directly through AI Studio is therefore usually not the right path.

Structural Lag in EU Regions

New Gemini models arrive first on the global endpoint and reach the EU multi-region endpoint with some delay in places; a dedicated EU single region (Frankfurt) is so far only offered for Gemini 3.5 Flash, while newer models (3.6, 3.7 and 3.8 Flash, Gemini 4 Argon) run in the EU only via the EU multi-region. Gemini 4 Argon, however, was available in the eu endpoint from its first GA day and additionally has single regions in London (UK) and Singapore. Developers also discuss this rollout pattern in Google’s own developer forum. For architecture planning that means: if you need a dedicated EU single-region endpoint, do not assume you can always use the newest model, and drive model selection through configuration rather than hard-wiring it in code.

Model lifecycle and deprecations

Google retires endpoints at a noticeable cadence. For anyone running production deployments, tracking the lifecycle belongs in the operational routine.

Shut down or superseded (dates on the Gemini API page are earliest possible dates):

• gemini-2.0-flash and gemini-2.0-flash-lite – 1 June 2026
• gemini-3-pro-preview – 9 March 2026
• gemini-3.1-flash-image-preview and gemini-3-pro-image-preview – 25 June 2026 (successors gemini-nano-banana-2.1 and gemini-3-pro-image respectively)
• Veo 2.0 and Veo 3.0 endpoints – 30 June 2026 (Gemini API and Agent Platform)
• imagen-4.0-generate-001 including its Ultra and Fast variants – 30 June 2026 on the Agent Platform, 17 August 2026 in the Gemini API (successor gemini-nano-banana-2.1)
• Grok 4.1 models on the Agent Platform – 20 August 2026
• gemini-robotics-er-1.6-preview – 31 August 2026 (successor gemini-robotics-er-2-preview)
• antigravity-preview-05-2026 – 5 October 2026 (successor antigravity-preview-09-2026)
• gemini-2.5-pro, gemini-2.5-flash and gemini-2.5-flash-lite – retirement on the Agent Platform 20 October 2026 (replacements gemini-3.8-flash, gemini-3.5-flash and gemini-3.5-flash-lite respectively); no shutdown date in the Gemini API, access limited to existing users
• Deprecated open model endpoints – 21 October 2026
• veo-3.1-generate-preview, veo-3.1-fast-generate-preview, veo-3.1-lite-generate-preview and gemini-omni-flash-preview – 22 October 2026 (successor gemini-omni-1.1-flash)
• deep-research-pro-preview-12-2025 – 23 October 2026 (successor deep-research-preview-04-2026 or deep-research-max-preview-04-2026)

Upcoming and deprecated:

• Gemini 3.6 Flash: retirement on the Agent Platform on 19 November 2026 (replacement gemini-3.8-flash); no shutdown date in the Gemini API.
• Gemini 3.7 Flash: deprecated in the Gemini API since 8 October 2026, requests are automatically routed to gemini-3.8-flash, no longer selectable in AI Studio. Retirement on the Agent Platform on 28 January 2027.
• Gemini 3.5 Flash: deprecated in the Gemini API since 8 October 2026, requests are automatically routed to gemini-3.6-flash. Still GA on the Agent Platform, retirement per the model page on 19 May 2027 or later.
• gemini-3.1-flash-tts-preview, gemini-3.1-flash-live-preview, gemini-2.5-flash-native-audio-preview-12-2025, gemini-2.5-flash-preview-tts and gemini-2.5-pro-preview-tts – 17 November 2026 (successors gemini-3.8-live, gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts)
• gemini-live-2.5-flash-native-audio (Agent Platform) – 13 December 2026 (successor gemini-3.8-live)
• gemini-2.5-flash-image (Nano Banana) – 15 March 2027 (successor gemini-3.1-flash-lite-image)
• gemini-3.1-flash-lite – Gemini API: shutdown no earlier than 7 May 2027 (successor gemini-3.5-flash-lite); Agent Platform: retirement 7 May 2027 or later
• gemini-embedding-001 – shutdown 14 May 2028 (successor gemini-embedding-2)
• gemini-3.1-flash-image (Nano Banana 2) – deprecated, no shutdown date (successor gemini-nano-banana-2.1)

For current models, the Agent Platform version overview lists retirement dates at least twelve months after release: Gemini 4 Argon 13 October 2027, 3.8 Flash 2 September 2027, 3.5 Flash-Lite 21 July 2027, each “or later”.

Parameter deprecation: since 21 July 2026, the sampling parameters temperature, top_p and top_k are deprecated (on 3.8 Flash additionally candidate_count). Gemini 4 Argon ignores the sampling parameters and rejects candidate_count, frequency_penalty, presence_penalty, thinking_budget and MINIMAL with HTTP 400. We recommend configuring model IDs and request parameters centrally rather than spreading them through application code.

EU availability

On the model pages of the Gemini Enterprise Agent Platform, Google documents data residency at rest and machine learning processing in the EU multi-region endpoint for seven current Gemini models: Gemini 4 Argon, 3.8 Flash, 3.7 Flash, 3.6 Flash, 3.5 Flash-Lite, 3.5 Flash and 3.1 Flash-Lite. Of these, only Gemini 3.5 Flash additionally has a single-region endpoint in the EU (Frankfurt, europe-west3), plus London (europe-west2, UK). Gemini 4 Argon is available as a single region only in London (UK) and Singapore. Gemini 3.1 Flash Image (Nano Banana 2) has been GA in the EU/US multi-region endpoint since 31 August 2026.

What works in the EU today

• Gemini 4 Argon via the EU multi-region endpoint with DRZ and MLP in the EU, including Provisioned Throughput, batch and Priority PayGo.
• Gemini 3.8 Flash, 3.7 Flash, 3.6 Flash, 3.5 Flash-Lite, 3.5 Flash and 3.1 Flash-Lite via the EU multi-region endpoint with DRZ and MLP in the EU.
• Supervised fine-tuning for Gemini 3.5 Flash, 3.5 Flash-Lite and 3.1 Flash-Lite (GA since 19 October 2026), including serving of the tuned models, in the EU multi-region endpoint with CMEK.
• Gemini 3.5 Flash additionally via single-region endpoints in Frankfurt and London (UK), the only current Flash generation with dedicated EU single-region access.
• Gemini 2.5 Pro via EU single regions (Belgium, Netherlands, Finland, Warsaw, Madrid, Milan, Paris; not Frankfurt or London), without an EU multi-region endpoint; its retirement date on the Agent Platform (20 October 2026) has passed.

What is still missing

• EU single regions for Gemini 4 Argon, 3.8 Flash, 3.7 Flash, 3.6 Flash, 3.5 Flash-Lite and 3.1 Flash-Lite (in the EU only the EU multi-region).
• Netherlands (europe-west4) for Gemini 3.5 Flash.
• An eu endpoint for Gemini 4 Argon Cyber Permissive (global and us only).
• Gemini 3.5 Pro (not released): we will assess EU residency at launch.

We recommend checking the model pages and the locations overview of the Gemini Enterprise Agent Platform regularly, as Google keeps extending regions and residency commitments.

A note on the EU AI Act

Enforcement of the EU AI Act transparency obligations has applied since 2 August 2026, with fines of up to 3 percent of worldwide annual turnover. According to the secondary sources available to us, Google has signed the EU AI Act Transparency Code of Practice and labels generated content with SynthID watermarks. This is an additional check in model selection but does not replace the residency assessment.

Integration with CompanyGPT

Gemini models are integrated in CompanyGPT. Gemini 3.5 Flash is wired in via the EU multi-region endpoint for GDPR-compliant reasoning and agentic coding workloads, with data residency at rest and machine learning processing in the EU. Gemini 2.5 Pro was wired in as a legacy option with DRZ and MLP in several EU single regions; since the Agent Platform lists its retirement for 20 October 2026, it is no longer an option for new projects. Since Google now also documents EU multi-region residency for Gemini 4 Argon, 3.8 and 3.6 Flash as well as 3.5 Flash-Lite, we are evaluating adding these models as further GDPR-compliant options.

Our recommendation

With Gemini 4 Argon, Google has for the first time made a frontier model available in the EU multi-region endpoint with DRZ and MLP from its first GA day. In EU projects the deciding factor is therefore less often “EU residency yes/no” than the task, the cost per task and whether a dedicated single-region endpoint is required:

• Demanding coding, analysis and agent workloads: Gemini 4 Argon via the EU multi-region endpoint, for example for multi-step code changes, long document and video analysis or workflows with very long outputs. Factor in the price doubling to $4/$20 from 1 December 2026 and the shared context for input, thinking and output. Check your own prompts and parameters against the migration guide before switching.
• Default model for most tasks: Gemini 3.8 Flash via the EU multi-region endpoint. Plan for roughly 30% higher token consumption per task compared with 3.7 Flash (Artificial Analysis measurement) and the price doubling from 1 January 2027. Plan Gemini 3.6 Flash only with its retirement on the Agent Platform on 19 November 2026 in mind.
• Existing integrations on 3.6 Flash, 3.7 Flash or 3.5 Flash: in the Gemini API, requests to 3.7 and 3.5 Flash are already routed automatically to 3.8 or 3.6 Flash. On the Agent Platform, complete the migration before the retirement dates: 3.6 Flash on 19 November 2026, 3.7 Flash on 28 January 2027.
• A dedicated single-region endpoint is mandatory (e.g. Frankfurt required by contract): Gemini 3.5 Flash on the Agent Platform, retirement per the model page no earlier than 19 May 2027.
• High-volume and latency-critical: Gemini 3.5 Flash-Lite ($0.30/1M input, 350 tokens/s, EU multi-region residency) or Gemini 3.1 Flash-Lite ($0.25/1M input, also with EU multi-region residency; Gemini API shutdown no earlier than 7 May 2027).
• Your own fine-tuning with EU residency: Gemini 3.5 Flash, 3.5 Flash-Lite or 3.1 Flash-Lite via supervised fine-tuning in the EU multi-region endpoint; tuned endpoints there cost 1.5 times the base model.
• Legacy estate with an EU single region outside Frankfurt: Gemini 2.5 Pro, technically surpassed; its retirement date on the Agent Platform (20 October 2026) has passed, so complete the migration to Gemini 3.8 Flash or 3.5 Flash-Lite.
• Cross-cloud alternative: Anthropic Claude (model page) via AWS Bedrock with EU geo inference, useful for multi-cloud strategies or when the workload is not bound to Google Cloud.

Unsure which option fits your data protection impact assessment? We review the residency and contractual situation in your specific project.