[ SUPPORT ]
Customer support
Email [email protected]. Staffed hours are Monday through Friday, 9:00 a.m. to 5:00 p.m. Pacific Time, excluding U.S. federal holidays. Email intake is available at all times.
These are acknowledgement and triage targets, not resolution guarantees. Include the affected model profile, Google Cloud region, impact, start time, request ID, and sanitized logs. Do not email credentials, tokens, prompts, outputs, or personal data.
[ SECURITY ]
Report a vulnerability
Send suspected vulnerabilities or product security incidents to [email protected]. Include the affected version or image digest, reproduction steps, impact, and a safe contact method. Do not include customer data or secrets that are not necessary to reproduce the issue.
We target acknowledgement within two business days and coordinate disclosure after remediation. Please do not disclose an unresolved report publicly.
[ PRIVACY ]
Privacy and data boundary
This page is static. Forms on renice.ai prepare a draft in the visitor's email client and do not submit form contents to a Renice AI backend. Information sent voluntarily by email is used to evaluate and respond to the request and is accessible to the applicable internal responder group.
The serving runtime operates in the customer's Google Cloud project. Prompts and outputs are not routinely sent to Renice AI. Publisher-hosted storage contains immutable model checkpoints and signed manifests, not customer prompts or outputs. Refer to the Marketplace listing for applicable commercial privacy terms.
[ TERMS ]
Product terms
Review the current terms and pricing in the Marketplace listing before deploying. Use is governed by the applicable Google Cloud Marketplace order, product terms and upstream licenses identified below.
The offering may not be used to cause or directly facilitate injury, conduct surveillance that violates internationally accepted norms, or contravene international law or human rights. Customers remain responsible for authorization, input governance, output review, and compliance with applicable law.
[ MODEL CARDS ]
Initial model profiles
Choose a complete model and hardware profile from the release you are installing. Available versions and their configuration are shown in the Marketplace listing.
Prompt limits include chat-template tokens. Your deployment's /v1/models endpoint reports the configured limit as max_prompt_length; see the installation guide's first API request.
GPT-OSS 120B
- Upstream
- openai/gpt-oss-120b
- Revision
b5c939de8f754692c1647ca79fbf85e8c1e70f8a- License
- Apache 2.0
- Marketplace profile
- Text, chat, completions, and streaming on one Cloud TPU v6e-8 slice using MXFP4.
- Limits
- It preserves the upstream Harmony response format and does not expose chain-of-thought to end users.
DeepSeek V4.1 Flash
- Upstream
- deepseek-ai/DeepSeek-V4.1-Flash
- Revision
dba1be0a40aa45a94ad051997016db3960a90277- License
- MIT
- Marketplace profile
- Text, chat, completions, and streaming on one Cloud TPU v5p slice with four physical chips in a 2x2x1 topology using native MXFP4.
- Limits
- It requires the product's dedicated chat adapter.
Qwen3.5 0.8B
- Upstream
- Qwen/Qwen3.5-0.8B
- Revision
2fc06364715b967f1860aea9cf38778875588b17- License
- Apache 2.0
- Marketplace profile
- Text, chat, completions, and streaming on a GKE Standard CPU profile using Int8.
- Limits
- Although the upstream model is multimodal, this product profile does not accept image input.
Common safety and use limitations
- Generated content can be inaccurate, biased, incomplete, unsafe, or inconsistent.
- Outputs require human review before high-impact or consequential use.
- Customers control access, inputs, retention, monitoring, and downstream use.
- No model training occurs inside the Marketplace deployment.
- Use the model and hardware combinations configured by your installed release. Availability in the selector does not establish performance or readiness for your workload.