Data residency

Data stored at rest in the customer selected location remains at rest in that location, independent of the Agent Platform endpoint called by that customer's request.

Where your data lives and is processed

Gemini Enterprise Agent Platform provides transparency on where your data is stored ("at rest") and where the actual model computation ("ML processing") happens.

1. Data-at-rest (storage)

When you store data on Agent Platform (such as custom model weights or metadata), it remains physically stored in the specific Google Cloud location you chose. This residency is maintained regardless of which endpoint you use to call the model.

2. ML processing (in-use)

ML processing is the method by which data is processed such that it produces model weights (tuning & training) or applies the model to the data for model inference. The geographic location of this processing is determined by your choice of endpoint:

  • Jurisdictional multi-region endpoints: When you use jurisdictional endpoints, ML processing stays within that specific geographical region (such as the United States or the European Union).

    • Operational efficiency (unified capacity): Multi-region endpoints eliminate regional resource siloes. Instead of purchasing and managing separate Provisioned Throughput allocations for individual regions (such as us-central1 and us-east4), you can deploy a single Provisioned Throughput commitment that covers your entire jurisdictional boundary (such as the United States).

    • Compliance enablement: Multi-region endpoints provide the data-in-use isolation controls required to help satisfy stringent regulatory frameworks. While multi-region endpoints serve as a foundational technical capability for complying with Department of Defense (DoD) Impact Level 5 (IL5) or International Traffic in Arms Regulations (ITAR) requirements, fully achieving compliance also requires broader customer-managed architecture, identity, and access controls.

  • Locational endpoints: These endpoints (like us-central1, europe-west1) ensure that ML processing remains entirely within the broader multi-regional or country jurisdiction associated with that region (for example, requests to us-central1 are processed within the United States).

    For European regions, local in-country processing varies by model type (for example, specific models may have local processing in France or Germany, while others are processed within the broader EU multi-region boundary). See the following section for the exact ML processing commitments for each model and location.

    • Compliance alignment: While locational endpoints meet standard enterprise data governance and sovereign requirements (such as GDPR and HIPAA), workloads requiring specialized government isolation frameworks like DoD IL5 or ITAR should be deployed on jurisdictional endpoints as part of a broader compliant architecture.

  • Global endpoints: Global endpoints route and process data anywhere globally, without restricting it to a specific geographic region. These endpoints (like https://aiplatform.googleapis.com) don't specify a region in the hostname. They are designed to maximize availability and minimize latency by terminating TLS sessions as close to the client as possible, but they don't provide regional isolation or data residency guarantees.

Supported models

The tables in this section cover the regional support for ML processing at a per-model basis for the following model categories:

Google models

To learn what capabilities support data residency, see Supported capabilities.

Model US multi-region EU multi-region Brazil(southamerica-east1) Canada(northamerica-northeast1) France(europe-west9) Germany(europe-west3) Netherlands(europe-west4) United Kingdom(europe-west2) Australia(australia-southeast1) India(asia-south1) Japan(asia-northeast1) Singapore(asia-southeast1) South Korea(asia-northeast3)
Gemini 3.5 Flash(gemini-3.5-flash)
Gemini 3.1 Flash-Lite(gemini-3.1-flash-lite)
Gemini 2.5 Flash Live API Native Audio(gemini-live-2.5-flash-native-audio)
Gemini 2.5 Flash, 128k(gemini-2.5-flash)
Gemini 2.5 Flash, 1M(gemini-2.5-flash)
Gemini 2.5 Flash Image(gemini-2.5-flash-image)
Gemini 2.5 Flash-Lite(gemini-2.5-flash-lite)
Gemini 2.5 Pro, 1M(gemini-2.5-pro)
Gemini 2.5 Pro, 64k(gemini-2.5-pro)
Tuning for Gemini 2.5 Flash(gemini-2.5-flash)
Tuning for Gemini 2.5 Flash-Lite(gemini-2.5-flash-lite)
Tuning for Gemini 2.5 Pro(gemini-2.5-pro)
Gemini Embedding(gemini-embedding-001)
Gemini Embedding 2(gemini-embedding-2)
Chirp 2: Transcription(chirp_2)
Chirp 3: Transcription(chirp_3)
Chirp 3: HD Voices
Chirp 3: Instant Custom Voice
Embeddings for Multimodal
Embeddings for Text(text-embedding-004)
Embeddings for Text(text-embedding-005)
Embeddings for Text(text-multilingual-embedding-002)

Google Cloud partner model support

Model US multi-region EU multi-region Belgium(europe-west1) Netherlands(europe-west4) Singapore(asia-southeast1) Taiwan(asia-east1) Global
Anthropic's Claude Sonnet 5 on Google Cloud
Anthropic's Claude Fable 5 on Google Cloud
Anthropic's Claude Haiku 4.5 on Google Cloud
Anthropic's Claude Opus 4 on Google Cloud
Anthropic's Claude Opus 4.1 on Google Cloud
Anthropic's Claude Opus 4.5 on Google Cloud
Anthropic's Claude Opus 4.8 on Google Cloud
Anthropic's Claude Opus 4.7 on Google Cloud
Anthropic's Claude Opus 4.6 on Google Cloud
Anthropic's Claude Sonnet 4 on Google Cloud
Anthropic's Claude Sonnet 4.5 on Google Cloud
Anthropic's Claude Sonnet 4.6 on Google Cloud
Anthropic's Claude 3.5 Haiku on Google Cloud (deprecated)
Anthropic's Claude 3 Haiku on Google Cloud (deprecated)
Anthropic's Claude 3.7 Sonnet on Google Cloud (deprecated)
Codestral (24.05)
Codestral 2
Mistral Large (24.07)
Mistral Medium 3
Mistral OCR (25.05)
Mistral Small 3.1 (25.03)

Google Cloud open model support

Model US multi-region EU multi-region Singapore(asia-southeast1) Global
DeepSeek-OCR
DeepSeek R1 (0528)
DeepSeek-V3.1
DeepSeek-V3.2
Gemma 4 26B A4B IT
GLM 4.7
GLM 5
gpt-oss 120B
gpt-oss 20B
Kimi K2 Thinking
Llama 3.3 70B (Preview)
Llama 4 Maverick 17B-128E (Preview)
Llama 4 Scout 17B-16E (Preview)
MiniMax M2
Multilingual E5 Large
Multilingual E5 Small
Qwen3 235B
Qwen3 Coder
Qwen3-Next-80B Instruct
Qwen3-Next-80B Thinking

What's next