Free to start

Google Vertex AI

Try Google Vertex AI on us. Starter credit covers your first requests, then one key reaches all 14 Google Vertex AI models, with the privacy posture shown on every route.

Google Vertex AIgoogle-vertex

ZDR · TR-funded

All providers

These privacy labels describe Google Vertex AI, the upstream model provider. Confidential requires all three: verified provider compute, end-to-end encryption and explicit zero data retention. Account and billing metadata are outside the content-retention claim.

Provider	Google Vertex AI
API operator country	US Legal home of the API operator, not an inference-location guarantee.
Headquarters	Not verified
Routing status	Active
Provider website	https://cloud.google.com/vertex-ai
Models	14 public models
Credits routes	14
Zero data retention	yes Reviewed TR-funded routes only; check each model
Verified confidential inference	Not verified
Policy note	TrustedRouter's managed Vertex AI account is covered by contractual Zero Data Retention, but eligibility is reviewed per model and feature. Durable caching was disabled on October 29, 2026 and memory-only caching verified in four regions. Affected Gemini routes remain excluded from strict ZDR pending prior data expiration and model-specific review. Safety-retention and unreviewed routes are also excluded, even when described upstream as ZDR-aligned. This guarantee applies only to reviewed TrustedRouter-funded routes. TrustedRouter does not invoke Google Search or Maps grounding or Gemini Live session resumption on these routes. Google AI Studio is classified separately. Policy source
Strict ZDR and safety retention	Retention varies by model and feature. Durable caching was disabled on the managed project on October 29, 2026; four regional checks confirmed memory-only caching. Affected Gemini routes remain excluded from strict ZDR pending prior data expiration and model-specific review. Advanced AI and Google Frontier safety-retention routes are also excluded. Google's ZDR-aligned wording is not a no-retention guarantee. Unreviewed models do not inherit the managed-account ZDR contract. Use `provider.min_privacy="zdr"` to reject retained or unverified routes, including during fallback.

Inference locations

Provider API declarations are also available on each model's endpoint API. Direct chat and Responses replies separate advertised availability from serving location in usage.inference_location; the actual serving region remains unknown when not reported.

usage.provider_usage.gateway_region identifies the TrustedRouter gateway, not the provider's inference location. A value such as europe-west4 does not mean the model ran in Europe. usage.inference_location.serving_region remains null when the provider has not reported it. Older gateways may return the gateway-only field as provider_usage.region.

Where the upstream GPUs execute the model, not the provider's headquarters, API ingress, or data storage. ZDR and confidential inference do not by themselves restrict geography.

Declared countries or regions: Not verified
No inference-location declaration has been verified for this provider.
Can locations change?: Not verified. Do not assume that the serving location is fixed.
Provider region pinning: Not verified.
Pinning through TrustedRouter: No verified provider-specific location pin documented here.
Infrastructure declaration: Inference location not verified; no country-specific residency commitment.
Evidence: Not yet reviewed.

Provider selection alone is not a region pin. TrustedRouter's US provider filter selects company jurisdiction, not GPU location. A regional gateway hostname does not pin a downstream GPU. For a binding residency requirement, confirm the model, region, failover scope and enforced controls before sending sensitive data.

Measured performance

19 samples

Continuously sampled across Google Vertex AI's routed models: p50 TTFT, effective throughput, and success rate. Effective throughput uses provider-reported output tokens over complete request time. Unsupported route and probe-configuration rows are separated from provider downtime. No prompt or output content stored.

p50 TTFT	1634 ms
Effective throughput	55 tok/s n=2
Uptime	94.74%

Model	p50 TTFT	Effective throughput	Uptime	Config excluded	Availability samples
google/gemini-2.5-flash	1203 ms	—	100.00%	—	2
google/gemini-3.5-flash-lite	1236 ms	—	100.00%	—	1
anthropic/claude-sonnet-4.6	1253 ms	—	100.00%	—	2
google/gemini-3.6-flash	1532 ms	—	100.00%	—	2
google/gemini-3.5-flash	1634 ms	59 tok/s n=1	100.00%	—	3
google/gemini-4-argon	2555 ms	—	50.00%	—	2
google/gemini-3.8-flash	2637 ms	—	100.00%	—	4
google/gemini-3.1-pro-preview	3752 ms	51 tok/s n=1	100.00%	—	1
google/gemini-2.5-pro	3843 ms	—	100.00%	—	1
google/gemini-2.5-flash-lite	—	—	100.00%	—	1

Google Vertex AI performance history · Full provider & model leaderboard.

Provider models

Models served by Google Vertex AI.

Each row links to pricing, provider, benchmark, and API pages for the model.

Model	AI IQ	Context	Input	Cached input	Output
`anthropic/claude-sonnet-4.6` Anthropic: Claude Sonnet 4.6 ZDR - TR-funded	IQ 115#59	1,000,000	$3.165/1M	$0.3165/1M	$15.825/1M
`google/gemini-2.5-flash` Google: Gemini 2.5 Flash ZDR - TR-funded	—	1,048,576	$0.3165/1M	$0.03165/1M	$2.6375/1M
`google/gemini-2.5-flash-lite` Google: Gemini 2.5 Flash Lite ZDR - TR-funded	—	1,048,576	$0.1055/1M	$0.01055/1M	$0.422/1M
`google/gemini-2.5-pro` Google: Gemini 2.5 Pro ZDR - TR-funded benchmarks	IQ 99#130	1,048,576	$1.31875/1M	$0.131875/1M to $0.26375/1M	$10.55/1M
`google/gemini-3.1-flash-lite` Google: Gemini 3.1 Flash Lite Cache remediation - review pending	IQ 104#102	1,048,576	$0.26375/1M	$0.026375/1M	$1.5825/1M
`google/gemini-3.1-pro-preview` Google: Gemini 3.1 Pro Preview Cache remediation - review pending benchmarks	IQ 122#38	1,048,576	$2.11/1M	$0.211/1M to $0.422/1M	$12.66/1M
`google/gemini-3.5-flash` Google: Gemini 3.5 Flash Cache remediation - review pending	IQ 120#47	1,048,576	$1.5825/1M	$0.15825/1M	$9.495/1M
`google/gemini-3.5-flash-lite` Google: Gemini 3.5 Flash Lite Cache remediation - review pending	IQ 104#103	1,048,576	$0.3165/1M	$0.03165/1M	$2.6375/1M
`google/gemini-3.6-flash` Google: Gemini 3.6 Flash Cache remediation - review pending	IQ 121#44	1,048,576	$0.79125/1M	$0.079125/1M	$3.95625/1M
`google/gemini-3.7-flash` Google: Gemini 3.7 Flash Cache remediation - review pending	IQ 126#25	1,048,576	$0.79125/1M	$0.079125/1M	$3.95625/1M
`google/gemini-3.8-flash` Google: Gemini 3.8 Flash Cache remediation - review pending	IQ 129#17	1,048,576	$0.79125/1M	$0.079125/1M	$3.95625/1M
`google/gemini-4-argon` Google: Gemini 4 Argon Cache remediation - review pending	IQ 140#3	1,048,576	$2.11/1M	$0.1055/1M	$10.55/1M
`google/veo-3.1` Google Veo 3.1 Retention unverified - not strict ZDR	—	2,500	selected route	Not published	selected route
`google/veo-3.1-fast` Google Veo 3.1 Fast Retention unverified - not strict ZDR	—	2,500	selected route	Not published	selected route

Questions

Does Google Vertex AI have zero data retention?

Reviewed managed Google Vertex AI routes can qualify for zero data retention. Check the exact model's endpoint policy: safety-retention, durable-cache and unreviewed routes are excluded. BYOK does not inherit this classification. Use provider.min_privacy=zdr to enforce eligibility.

Is Google Vertex AI end-to-end encrypted?

TrustedRouter does not currently mark Google Vertex AI as end-to-end encrypted at the provider boundary. The TrustedRouter gateway is still attested, but the selected provider normally receives the request in order to run the model. Use trustedrouter/e2e for the stronger route requirement.

Which Google Vertex AI models are available through TrustedRouter?

This page currently lists 14 public Google Vertex AI models, with live pricing, route count, context length, measured performance when available, and links to each model's provider and benchmark pages.