{ "schemaVersion": 1, "reviewed": "2026-11-09", "status": "research-preview", "listings": [ { "id": "azure-gpt-4-1-mini-australia-east", "provider": "Azure", "initials": "AZ", "service": "Foundry · Standard · Australia East", "model": "gpt-4.1-mini · 2025-04-14", "scope": "Model configuration", "processing": "documented", "access": "Azure subscription and model deployment; check quota and live availability", "billing": "Pay per token; Australia East input/output price by SKU not yet captured", "commitment": "none", "summary": "The current regional availability matrix lists this model in Australia East for Standard/Regional usage billing.", "caveat": "Availability changes over time and is subscription-specific. Check live deployment access, quota and version lifecycle.", "storage": "Model processing is region-scoped in Australia East. Stored resource data and stateful features need their own location and retention review.", "retention": "Inputs and outputs are not used to train base models without permission. Abuse monitoring may retain flagged samples for human review unless eligible Modified Abuse Monitoring terms apply. Stateful features can store content.", "ownership": "Microsoft operates this Azure service. Australian processing does not establish Australian ownership.", "setup": [ "Create a Standard deployment in Australia East with version 2025-04-14.", "Avoid Global Standard and Data Zone deployments if the workload requires Australia-only inference.", "Check current price, quota, retirement status and stateful feature storage." ], "gaps": [ "Current Australia East price by SKU", "Live deployment availability for the intended subscription", "Feature-specific storage and abuse-monitoring terms" ], "evidence": [ { "claim": "Microsoft lists gpt-4.1-mini · 2025-04-14 for Standard/Regional deployment in Australia East. Standard/Regional prompts and responses are processed in the deployment region.", "url": "https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure-region-availability", "title": "Azure model and region availability", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Standard deployment is pay-per-token and region-scoped where available.", "url": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/deployment-types", "title": "Azure deployment types", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Training, stateful feature storage, content filters and abuse monitoring are distinct controls.", "url": "https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/openai/data-privacy", "title": "Azure OpenAI data privacy", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "azure-gpt-4o-australia-east", "provider": "Azure", "initials": "AZ", "service": "Foundry · Standard · Australia East", "model": "gpt-4o · 2024-11-20", "scope": "Model configuration", "processing": "documented", "access": "Azure subscription and model deployment; check quota and live availability", "billing": "Pay per token; Australia East input/output price by SKU not yet captured", "commitment": "none", "summary": "The current regional availability matrix lists gpt-4o 2024-11-20 in Australia East for Standard/Regional usage billing.", "caveat": "This is a dated model version and may be retired. Confirm current Foundry catalog availability before relying on it.", "storage": "Model processing is region-scoped in Australia East. Stored resource data and stateful features need their own location and retention review.", "retention": "Inputs and outputs are not used to train base models without permission. Abuse monitoring may retain flagged samples for human review unless eligible Modified Abuse Monitoring terms apply. Stateful features can store content.", "ownership": "Microsoft operates this Azure service. Australian processing does not establish Australian ownership.", "setup": [ "Create a Standard deployment in Australia East with version 2024-11-20.", "Avoid Global Standard and Data Zone deployments if the workload requires Australia-only inference.", "Check model retirement, price, quota and abuse-monitoring terms." ], "gaps": [ "Current Australia East price by SKU", "Live deployment availability and retirement status", "Feature-specific storage and abuse-monitoring terms" ], "evidence": [ { "claim": "Microsoft lists gpt-4o · 2024-11-20 for Standard/Regional deployment in Australia East. Standard/Regional prompts and responses are processed in the deployment region.", "url": "https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure-region-availability", "title": "Azure model and region availability", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Standard deployment is pay-per-token and region-scoped where available.", "url": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/deployment-types", "title": "Azure deployment types", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Training, stateful feature storage, content filters and abuse monitoring are distinct controls.", "url": "https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/openai/data-privacy", "title": "Azure OpenAI data privacy", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "azure-embedding-3-large-australia-east", "provider": "Azure", "initials": "AZ", "service": "Foundry · Standard · Australia East", "model": "text-embedding-3-large · version 1", "scope": "Model configuration", "processing": "documented", "access": "Azure subscription and model deployment; check quota and live availability", "billing": "Pay per token; Australia East price by SKU not yet captured", "commitment": "none", "summary": "The current regional availability matrix lists this embedding model in Australia East for Standard/Regional usage billing.", "caveat": "The embedding request is regional; vector index and source-document residency remain separate configuration decisions.", "storage": "Inference is region-scoped in Australia East. Embedding indexes and source documents have separate storage locations.", "retention": "Inputs and outputs are not used to train base models without permission. Abuse monitoring may retain flagged samples for human review unless eligible Modified Abuse Monitoring terms apply. Stateful features can store content.", "ownership": "Microsoft operates this Azure service. Australian processing does not establish Australian ownership.", "setup": [ "Create a Standard deployment in Australia East.", "Keep the vector database and source documents in an acceptable location.", "Check price, quota, supported dimensions and current endpoint identifier." ], "gaps": [ "Current Australia East price by SKU", "Vector database location for a full retrieval workflow", "Live deployment availability for the intended subscription" ], "evidence": [ { "claim": "Microsoft lists text-embedding-3-large · version 1 for Standard/Regional deployment in Australia East. Standard/Regional prompts and responses are processed in the deployment region.", "url": "https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure-region-availability", "title": "Azure model and region availability", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Standard deployment is pay-per-token and region-scoped where available.", "url": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/deployment-types", "title": "Azure deployment types", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Training, stateful feature storage, content filters and abuse monitoring are distinct controls.", "url": "https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/openai/data-privacy", "title": "Azure OpenAI data privacy", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "gcp-gemini-3-5-flash", "provider": "Google Cloud", "initials": "GC", "service": "Vertex AI · regional standard", "model": "Gemini 3.5 Flash · gemini-3.5-flash", "scope": "Model configuration", "processing": "documented", "access": "Google Cloud project, billing and Vertex AI API enabled; check quota", "billing": "USD per 1M tokens: $1.65 input / $9.90 text output; cached input $0.165", "commitment": "none", "summary": "Google’s per-model processing table documents Australia ML processing for this model. Use Sydney regional endpoint australia-southeast1.", "caveat": "A global endpoint may process anywhere. Storage controls and model processing location are separate; use the non-global API path.", "storage": "Model inference executes in Australia on supported regional endpoint. Stored data follows separately configured Google Cloud locations.", "retention": "Review logging, caching, grounding and feature-specific policy separately; this record does not claim zero retention.", "ownership": "Google Cloud operates this service. Australian processing does not establish Australian ownership.", "setup": [ "Use australia-southeast1 regional endpoint, not global.", "Check Google’s model residency table and regional pricing at deployment time.", "Review prompt logging and any connected grounding/vector-store service." ], "gaps": [ "Subscription-specific quota and enablement", "Feature-level retention conditions", "Independent operational test not performed" ], "evidence": [ { "claim": "The model’s Australian entry in the per-model ML processing table documents processing in Australia.", "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/resources/data-residency", "title": "Data residency by model", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Google’s model page lists regional invocation in australia-southeast1.", "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/3-5-flash", "title": "Gemini 3.5 Flash", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Published non-global rates are $1.65 input and $9.90 text output per 1M tokens, with $0.165 cached input.", "url": "https://cloud.google.com/model-garden/pricing", "title": "Gemini API pricing", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "gcp-gemini-3-5-flash-lite", "provider": "Google Cloud", "initials": "GC", "service": "Vertex AI · regional standard", "model": "Gemini 3.5 Flash-Lite · gemini-3.5-flash-lite", "scope": "Model configuration", "processing": "documented", "access": "Google Cloud project, billing and Vertex AI API enabled; check quota", "billing": "USD per 1M tokens: $0.33 input / $2.75 text output; cached input $0.033", "commitment": "none", "summary": "Google’s per-model processing table documents Australia ML processing for this model. Use Sydney regional endpoint australia-southeast1.", "caveat": "Global endpoints may process anywhere. Storage controls and model processing are separate; use the non-global API path.", "storage": "Model inference executes in Australia on supported regional endpoint. Stored data follows separately configured Google Cloud locations.", "retention": "Review logging, caching, grounding and feature-specific policy separately; this record does not claim zero retention.", "ownership": "Google Cloud operates this service. Australian processing does not establish Australian ownership.", "setup": [ "Use australia-southeast1 regional endpoint, not global.", "Check Google’s model residency table and regional pricing at deployment time.", "Review prompt logging and any connected grounding/vector-store service." ], "gaps": [ "Subscription-specific quota and enablement", "Feature-level retention conditions", "Independent operational test not performed" ], "evidence": [ { "claim": "The model’s Australian entry in the per-model ML processing table documents processing in Australia.", "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/resources/data-residency", "title": "Data residency by model", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Google’s model page lists regional invocation in australia-southeast1.", "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/3-5-flash-lite", "title": "Gemini 3.5 Flash-Lite", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Published non-global rates are $0.33 input and $2.75 text output per 1M tokens, with $0.033 cached input.", "url": "https://cloud.google.com/model-garden/pricing", "title": "Gemini API pricing", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "aws-deepseek-v3-2-sydney", "provider": "AWS", "initials": "AWS", "service": "Bedrock · on-demand · ap-southeast-2", "model": "DeepSeek V3.2 · deepseek.v3.2", "scope": "Model configuration", "processing": "documented", "access": "AWS account; model access and quota required", "billing": "Standard on-demand; $0.6386 input / $1.9055 output per 1M tokens (AWS page shows $, currency not explicit)", "commitment": "none", "summary": "Invoke the model in the Sydney Bedrock region through its regional endpoint. AWS lists standard on-demand token pricing.", "caveat": "Use the regional model identifier and endpoint directly. Cross-region inference profiles have a different and changeable routing boundary.", "storage": "Prompts and outputs are processed in the selected AWS region. Bedrock retention is configured separately and varies by model.", "retention": "Check model-specific abuse-detection requirements. Zero retention is configurable for supported models and is not assumed by this entry.", "ownership": "AWS operates Bedrock infrastructure; DeepSeek is the model developer.", "setup": [ "Call ap-southeast-2 directly using deepseek.v3.2.", "Do not replace the regional model ID with an AU geographic or global profile.", "Review Bedrock retention settings and DeepSeek model terms." ], "gaps": [ "Confirm account-specific enablement and rate limits", "AWS page does not make billing currency explicit in the captured price table", "Independent operational test not performed" ], "evidence": [ { "claim": "DeepSeek V3.2 has an in-region Sydney model identifier and model card.", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-deepseek-deepseek-v3-2.html", "title": "DeepSeek V3.2 model card", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "AWS lists this model as available in ap-southeast-2 with direct in-region invocation.", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/models-region-compatibility.html", "title": "Model support by AWS Region", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "The AWS pricing table lists Sydney Standard-tier token prices; currency symbol is shown without an explicit currency label in the captured entry.", "url": "https://aws.amazon.com/bedrock/pricing/", "title": "Amazon Bedrock pricing", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Bedrock data retention is a separate account/model control; do not assume zero retention.", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/data-retention.html", "title": "Data retention controls", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "aws-nova-lite-sydney", "provider": "AWS", "initials": "AWS", "service": "Bedrock · on-demand · ap-southeast-2", "model": "Amazon Nova Lite · amazon.nova-lite-v1:0", "scope": "Model configuration", "processing": "documented", "access": "AWS account; model access and quota required", "billing": "Standard on-demand per-token; check current Sydney price", "commitment": "none", "summary": "Amazon Nova Lite supports direct in-region inference in Sydney using Bedrock on-demand access.", "caveat": "Do not use Global inference. Nova Lite also supports an Australian cross-region profile, which can have different eligible destinations.", "storage": "Prompts and outputs are processed in the selected AWS region. Retained content depends on Bedrock model settings.", "retention": "Review current model-specific retention/abuse policy and account settings.", "ownership": "Amazon develops Nova and operates Bedrock infrastructure.", "setup": [ "Invoke amazon.nova-lite-v1:0 directly in ap-southeast-2.", "Review latest Sydney pricing and the model-specific retention setting." ], "gaps": [ "Numeric Sydney token pricing", "Account-specific enablement and quotas", "Independent operational test not performed" ], "evidence": [ { "claim": "Amazon Nova Lite is available for direct regional Bedrock invocation and has model identifier amazon.nova-lite-v1:0.", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-amazon-nova-lite.html", "title": "Amazon Nova Lite model card", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "The Sydney region model matrix lists regional support for Nova Lite.", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/models-region-compatibility.html", "title": "Model support by AWS Region", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Bedrock retention settings vary by model and configuration.", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/data-retention.html", "title": "Data retention controls", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "databricks-gpt-oss-20b-sydney", "provider": "Databricks", "initials": "DB", "service": "Foundation Model APIs · pay-per-token · ap-southeast-2", "model": "databricks-gpt-oss-20b · system.ai.gpt-oss-20b", "scope": "Model configuration", "processing": "conditional", "access": "Databricks workspace in supported region and API token", "billing": "Pay-per-token; no capacity commitment; per-model price not publicly specified", "commitment": "none", "summary": "The model is listed for Databricks-hosted inference in AWS Sydney with a pay-per-token endpoint. With cross-Geo disabled, processing is bounded to the Australia and New Zealand Geo.", "caveat": "This is an Australia–New Zealand geography, not Australia-only. Cross-Geo is enabled by default outside the US and EU unless the workspace has a compliance security profile.", "storage": "Databricks data processing follows the workspace Geo for designated services when cross-Geo is disabled. Australia and New Zealand form one Geo.", "retention": "Model-specific retention and Workspace logging settings still need review.", "ownership": "Databricks hosts the model. Sydney location does not establish Australian ownership or infrastructure operator domicile.", "setup": [ "Use the Databricks pay-per-token endpoint in an ap-southeast-2 workspace.", "Have an account admin disable cross-Geo processing; confirm any compliance security profile defaults.", "Assess whether Australia and New Zealand as one processing boundary is acceptable." ], "gaps": [ "No public model-specific input/output rates found", "AU/NZ Geo may not meet strict Australia-only requirements", "Check workspace-tier quotas and retention settings" ], "evidence": [ { "claim": "The Databricks-supported model list identifies the service name and indicates Databricks hosting unless a third-party Hosting field is shown.", "url": "https://docs.databricks.com/aws/en/machine-learning/foundation-model-apis/supported-models", "title": "Supported Foundation Model API models", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "The Sydney region table lists databricks-gpt-oss-20b for pay-per-token use.", "url": "https://docs.databricks.com/aws/en/unity-gateway/model-region-availability", "title": "Databricks model availability by region", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Australia and New Zealand form one Databricks Geo; cross-Geo processing is enabled by default for many non-US/EU workspaces.", "url": "https://docs.databricks.com/aws/en/resources/databricks-geos", "title": "Databricks Geos and data residency", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "databricks-gemma-3-12b-sydney", "provider": "Databricks", "initials": "DB", "service": "Foundation Model APIs · pay-per-token · ap-southeast-2", "model": "databricks-gemma-3-12b · system.ai.gemma-3-12b", "scope": "Model configuration", "processing": "conditional", "access": "Databricks workspace in supported region and API token", "billing": "Pay-per-token; no capacity commitment; per-model price not publicly specified", "commitment": "none", "summary": "Databricks lists a hosted Gemma endpoint in AWS Sydney for pay-per-token use. When cross-Geo is disabled, the processing boundary is Australia and New Zealand together.", "caveat": "This is not an Australia-only Geo. Cross-Geo processing is enabled by default for non-US/EU workspaces without a compliance security profile.", "storage": "Databricks data processing follows the workspace Geo for designated services when cross-Geo is disabled. Australia and New Zealand form one Geo.", "retention": "Model-specific retention and workspace logging settings require review.", "ownership": "Databricks hosts the model. Australian infrastructure location does not establish Australian corporate ownership.", "setup": [ "Use the Databricks pay-per-token endpoint in an ap-southeast-2 workspace.", "Have an account admin disable cross-Geo processing.", "Check current quota, features and whether the AU/NZ boundary is acceptable." ], "gaps": [ "No public model-specific input/output rates found", "Strict Australia-only processing is not established by this Geo", "Workspace-tier quota and retention details" ], "evidence": [ { "claim": "The supported model list gives the Databricks endpoint and model service name.", "url": "https://docs.databricks.com/aws/en/machine-learning/foundation-model-apis/supported-models", "title": "Supported Foundation Model API models", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "The Sydney region table lists databricks-gemma-3-12b for pay-per-token use.", "url": "https://docs.databricks.com/aws/en/unity-gateway/model-region-availability", "title": "Databricks model availability by region", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Cross-Geo disabled bounds designated-service processing to the workspace Geo, which groups Australia and New Zealand.", "url": "https://docs.databricks.com/aws/en/resources/databricks-geos", "title": "Databricks Geos and data residency", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "govai-model-brokerage", "provider": "GovAI", "initials": "GA", "service": "Model Brokerage · Australian Government", "model": "GPT-4o onshore instance; Claude and Nova options subject to agreement", "scope": "Service profile", "processing": "conditional", "access": "Australian Government agency only; GovTEAMS, senior accountable official, onboarding and IP allowlist", "billing": "Pilot pay-as-you-go by model use; no upfront costs during pilot", "commitment": "none", "summary": "A government-managed API gateway for eligible Australian Government agencies, with specific onshore model arrangements.", "caveat": "The published model list does not map every model/version to its onshore route. Agency-only access and onboarding exclude general commercial self-service.", "storage": "The service claims onshore routing for specific models; other options may support global inference.", "retention": "Page says underlying providers do not retain or train on data and only request metadata is logged by default; verify contract and model route.", "ownership": "Australian Government-managed brokerage service; underlying inference providers vary by selected model.", "setup": [ "Confirm agency eligibility and request the Model Brokerage guide.", "Obtain the route and processing conditions for the exact model/version.", "Register application IPs and provision access through the GovAI Developer Portal." ], "gaps": [ "Full service guide and model-specific onshore routing table", "Current pilot status, agency pricing schedule and contract terms", "Independent assessment scope for each downstream model path" ], "evidence": [ { "claim": "The brokerage is restricted to Australian Government agencies and exposes an API gateway to approved model providers.", "url": "https://www.govai.gov.au/govai-model-brokerage", "title": "GovAI Model Brokerage", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "The page lists an onshore GPT-4o instance and Claude/Nova models, while noting versions may lag.", "url": "https://www.govai.gov.au/govai-model-brokerage", "title": "Available models and regional conditions", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "During the pilot, the agency pays for model use with no upfront costs; account onboarding is required.", "url": "https://www.govai.gov.au/govai-model-brokerage", "title": "Access and pricing", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "azure-standard", "provider": "Azure", "initials": "AZ", "service": "Foundry · Standard", "model": "Model/version selection pending", "scope": "Service profile", "processing": "conditional", "access": "Azure account; deployment and quota required", "billing": "Pay per token", "commitment": "none", "summary": "Geography-based processing with usage billing. The exact model and deployment type matter.", "caveat": "An Australian resource using Global Standard or an APAC Data Zone is not an Australia-only configuration.", "storage": "Stored data remains in the designated Azure geography under the deployment documentation.", "retention": "Abuse monitoring, feature storage and retention require a separate review.", "ownership": "Microsoft; Australian hosting does not establish Australian ownership or operational control.", "setup": [ "Choose Standard, not Global Standard or Data Zone Standard.", "Check the exact model/version is available in the chosen Australian geography.", "Review retention, feature exceptions and quota before sending production data." ], "gaps": [ "Exact Australian model/version availability", "Retention and support access conditions", "Current regional prices" ], "evidence": [ { "claim": "Standard processing is bounded to the selected Azure geography; Standard billing is pay per token.", "url": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/deployment-types", "title": "Deployment types", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "aws-bedrock", "provider": "AWS", "initials": "AWS", "service": "Bedrock · on-demand", "model": "Model/profile selection pending", "scope": "Service profile", "processing": "conditional", "access": "AWS account; model access and quota required", "billing": "On-demand usage", "commitment": "none", "summary": "Model access through Bedrock, with processing boundaries that depend on the invocation route.", "caveat": "An Australian source region does not prove Australian inference. Inspect every destination in the selected inference profile.", "storage": "Abuse-detection storage can occur in a destination region for relevant models.", "retention": "Model-specific abuse-detection requirements need review.", "ownership": "Amazon Web Services; local infrastructure and ownership are separate dimensions.", "setup": [ "Select the exact model ID and invocation method.", "Check all eligible destination regions, including failover.", "Review model-specific logging and abuse detection." ], "gaps": [ "Exact model and destination set", "Australian-only on-demand availability", "Current model-specific pricing" ], "evidence": [ { "claim": "Geographic inference profiles can move prompts and outputs outside the source region.", "url": "https://docs.aws.amazon.com/bedrock/latest/userguide/geographic-cross-region-inference.html", "title": "Geographic cross-region inference", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "google-cloud", "provider": "Google Cloud", "initials": "GC", "service": "Managed generative AI APIs", "model": "Model/endpoint selection pending", "scope": "Service profile", "processing": "unverified", "access": "Google Cloud project and model quota", "billing": "Model-specific; pricing review pending", "commitment": "unknown", "summary": "Evaluate the model’s processing commitment separately from its endpoint location.", "caveat": "A regional endpoint alone does not guarantee in-region ML processing.", "storage": "Storage location and inference processing are separate controls.", "retention": "Feature-specific logging, caching and retention not yet reviewed.", "ownership": "Google Cloud; Australian endpoint availability does not establish local ownership.", "setup": [ "Select the exact model, version and endpoint.", "Read the processing commitment applicable to that model.", "Check features such as grounding and caching separately." ], "gaps": [ "Australian processing commitment for selected model", "Usage pricing and minimum requirements", "Feature-level data handling" ], "evidence": [ { "claim": "Endpoint location alone does not guarantee data residency or in-region ML processing.", "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/resources/locations", "title": "Deployments and endpoints", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Storage and model computation have distinct location controls.", "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/resources/data-residency", "title": "Data residency", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "openai-australia", "provider": "OpenAI", "initials": "OA", "service": "API · Australian data residency", "model": "Eligible models and endpoints only", "scope": "Service profile", "processing": "offshore", "access": "Eligibility approval and amended retention agreement", "billing": "Usage billing; residency pricing conditions apply", "commitment": "unknown", "summary": "Australian storage for eligible API projects, without an Australian inference processing guarantee.", "caveat": "Australian storage is not Australian inference. The documented Australian region does not support regional processing.", "storage": "Eligible customer content at rest in Australia; system data excluded.", "retention": "Non-US residency requires approval for abuse-monitoring controls and a Modified Retention amendment.", "ownership": "OpenAI; eligibility and service terms must be checked for the contracting entity.", "setup": [ "Confirm residency eligibility with OpenAI.", "Create a project with Australian residency and use supported endpoints.", "Do not use this option where Australian inference processing is required." ], "gaps": [ "Account eligibility and commercial prerequisites", "Endpoint and model coverage", "Feature-specific retention exceptions" ], "evidence": [ { "claim": "Australia supports regional storage but not regional processing; eligible customers need approval for abuse-monitoring controls and a Modified Retention amendment.", "url": "https://developers.openai.com/api/docs/guides/your-data", "title": "Data controls in the OpenAI platform", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Default abuse-monitoring logs can retain customer content for up to 30 days; Modified Abuse Monitoring and Zero Data Retention require approval and have exceptions.", "url": "https://developers.openai.com/api/docs/guides/your-data", "title": "Data retention controls", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "databricks-pay-per-token", "provider": "Databricks", "initials": "DB", "service": "Foundation Model APIs", "model": "Sydney options include databricks-gpt-oss-20b and databricks-gemma-3-12b", "scope": "Service profile", "processing": "conditional", "access": "Databricks workspace; entitlement review required", "billing": "Pay-per-token option", "commitment": "unknown", "summary": "The Australian workspace Geo includes Australia and New Zealand; cross-Geo settings and model hosting determine the actual boundary.", "caveat": "Australia and New Zealand share a Databricks Geo. Cross-Geo processing is enabled by default on many non-US/EU workspaces.", "storage": "Workspace location alone is insufficient; service and model terms require review.", "retention": "Model and workspace-level retention needs review. External-model APIs forward requests to their named provider.", "ownership": "Databricks plus applicable infrastructure or external model provider.", "setup": [ "Distinguish pay-per-token from provisioned and external-model endpoints.", "Check the processing geography for the exact model and workspace.", "Review workspace costs and model-provider exceptions." ], "gaps": [ "Australian processing for a named model", "Workspace fees and minimum requirements", "Retention and external model routing" ], "evidence": [ { "claim": "Processing depends on workspace and model and may occur outside the originating region or cloud.", "url": "https://docs.databricks.com/aws/en/machine-learning/foundation-model-apis/compliance", "title": "Foundation Model APIs compliance and security", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "toothfairyai", "provider": "ToothFairyAI", "initials": "TF", "service": "Sovereign LLM API · AU", "model": "19 AU-region model APIs advertised; named examples include mystica_15, TF Sorcerer and TF Mystica", "scope": "Service profile", "processing": "unverified", "access": "Self-service advertised", "billing": "Published per-token rates advertised", "commitment": "unknown", "summary": "A smaller provider advertising Australian regional inference and OpenAI-compatible APIs.", "caveat": "The provider advertises regional inference and zero retention. Its privacy policy separately allows platform/workspace data retention and overseas subprocessors.", "storage": "Australian regional service advertised; logs and backups not independently reviewed.", "retention": "Zero data retention advertised; exceptions and contractual scope pending.", "ownership": "Terms identify ToothFairyAI Pty Ltd (Australian company); the privacy policy says platform hosting uses AWS and permits overseas service providers.", "setup": [ "Choose the AU regional endpoint.", "Verify the exact model is served locally rather than routed elsewhere.", "Obtain retention, subprocessor and failover terms." ], "gaps": [ "Contractual processing boundary", "Ownership and infrastructure operator", "Exact live-model rates and minimum account funding" ], "evidence": [ { "claim": "The provider advertises AU endpoints, 19 Australian models, per-token pricing and zero prompt/output retention. These are provider claims, not independent location verification.", "url": "https://toothfairyai.com/en/sovereign", "title": "Sovereign LLM API", "kind": "Provider claim", "checked": "2026-11-09" }, { "claim": "Terms identify ToothFairyAI Pty Ltd (ABN 25 668 487 705; ACN 668 487 705) as the Australian contracting entity.", "url": "https://toothfairyai.com/en/terms-and-conditions", "title": "Terms and conditions", "kind": "Documentation", "checked": "2026-11-09" }, { "claim": "Privacy policy identifies AWS hosting, possible overseas service providers, and up to 30 days of workspace retention after subscription expiry.", "url": "https://toothfairyai.com/en/privacy-policy", "title": "Privacy policy", "kind": "Documentation", "checked": "2026-11-09" } ] }, { "id": "xinference", "provider": "Xinference", "initials": "XI", "service": "Model API · Australia", "model": "DeepSeek V4.1 Flash and GLM-5.3; names and prices should be checked against the live rate feed", "scope": "Service profile", "processing": "unverified", "access": "Self-service advertised", "billing": "Per-token. Model API page lists $0.30/$1.20 DeepSeek V4.1 Flash and $1.40/$4.40 GLM-5.3 per 1M; currency not labelled.", "commitment": "unknown", "summary": "A self-service, OpenAI-compatible API advertised as running in Australian data centres, with named DeepSeek and GLM models.", "caveat": "The provider’s homepage and Model API page list different model names/prices. Current Model API page names DeepSeek V4.1 Flash and GLM-5.3; verify its live rate feed before publishing numeric prices.", "storage": "Australian hosting advertised; storage categories not yet verified.", "retention": "Provider says it does not train on prompts/outputs. Request logs are inspectable; prompt/output retention periods and subprocessors are not published on the reviewed Model API page.", "ownership": "Australian processing and no-training statements are provider claims; reviewed public pages do not establish legal contracting entity, subprocessors, or failover destinations.", "setup": [ "Confirm you are purchasing the shared Model API.", "Confirm the exact model and Australian processing boundary.", "Review retention and minimum payment requirements." ], "gaps": [ "Service-specific terms and subprocessors", "Routing and failover destinations", "Current live price feed and billing currency" ], "evidence": [ { "claim": "The provider advertises Australian-hosted DeepSeek V4.1 Flash ($0.30 input / $1.20 output) and GLM-5.3 ($1.40 / $4.40) per 1M tokens. The page does not clearly state currency; rates are self-reported and can change.", "url": "https://xinference.co/model-api.html", "title": "Model API and published rates", "kind": "Provider claim", "checked": "2026-11-09" }, { "claim": "The provider says prompts and outputs are processed in Australian data centres and are not used for training absent written agreement.", "url": "https://xinference.co/model-api.html", "title": "Model API processing statement", "kind": "Provider claim", "checked": "2026-11-09" } ] } ] }