Key points
- The service routes requests across twelve live models using an OpenAI-compatible endpoint.
- You.com guarantees ninety-nine point nine percent uptime and processes data in memory only.
- Pricing is pay-as-you-go with no platform fee and automatic failover to US partners.
The short version
- You.com launched Inference Router, a managed API for running open-weight models in production environments.
- The service offers zero data retention, hundred percent US hosting and server-side web search capabilities.
- It provides a ninety-nine point nine percent uptime service level agreement with no platform fees for users.
The details
You.com announced the launch of its Inference Router, a service designed to help organizations run open-weight models in production. According to MarkTechPost, this new offering is an OpenAI-compatible API that routes requests across twelve live models. The service bundles a minimum uptime service level agreement of ninety-nine point nine percent and promises zero data retention. It also includes server-side web search and ensures all processing occurs within the United States. Users pay as they go, with no platform fee required. The available models range from GPT-OSS 120B to larger options like Kimi K3 and Qwen3.8 Max.
Kimi K3 is a 2.8 trillion-parameter model, while Qwen3.8 Max is a 2.4 trillion parameter mixture-of-experts model. The service is fully managed, meaning users do not need to manage GPUs themselves. Requests first hit You.com’s own inference capacity, named Inference Direct. If that capacity is down or full, the request fails over to vetted partners. The official post names Parasail, Wafer, Alibaba Cloud, Meta and Inception as partners, with Prime Intellect coming soon. All partners are required to be hundred percent US-hosted and must meet the zero data retention rules.
The context
Performance testing on You.com’s LiveNewsBench showed that open-weight models combined with You.com search matched frontier models at a lower cost. GLM-5.3 with You.com search scored eighty-six point five three percent at a cost of forty-eight point five cents per question. This was compared to ninety-two cents for Claude Sonnet 5.5, based on self-reported figures. Per-token prices for other models sit near market rates.
Kimi K3 costs two dollars eighty-five cents per input token and fourteen dollars twenty-five cents per output token. This is only five percent below Together AI’s pricing of three dollars and fifteen dollars respectively. The service uses one API for many open-weight models. The endpoint is api.you.com, with routes compatible with both OpenAI and Anthropic. Existing SDK code can move over by changing the base URL and key.
What to watch
The zero data retention policy states that prompts and outputs are processed in memory and not stored. This covers both model calls and web search requests. You.com keeps only operational metadata, such as token counts, timestamps and account IDs. Prompt hashes, de-identified text and disk-backed prompt caches are explicitly excluded from storage. Accounts run on SOC 2 compliant infrastructure. A data processing agreement is available on request.
You.com also offers spend caps per API key and daily usage views by model, cost and provider. Failover does not change the bill, as each model has one published per-token price. The You.com team frames the problem around stalled migrations, noting that dev teams often test open-weight models on aggregators but then hit security reviews. Reviewers typically ask about zero data retention, US hosting, access controls and data processing agreements.
What it means for organisations
Organizations should assess whether this service meets their data residency and privacy requirements. Teams usually need to verify that the zero data retention policy aligns with internal security standards. Check whether the US hosting requirement satisfies your legal constraints. Determine if the failover mechanism to partners introduces any new risk vectors. Discuss the cost structure with finance teams to ensure budget alignment.
Inform legal departments about the data processing agreement and compliance certifications. Ensure that development teams understand the migration path from existing providers. Evaluate if the supported models meet your performance and accuracy needs. Review access controls and API key management procedures. Confirm that the service level agreement matches your operational requirements. Communicate any changes to vendors or suppliers who rely on AI outputs.
Background: Data residency
Data residency laws require organizations to store specific types of information within defined geographic boundaries. You must identify where your cloud providers place physical servers. Failure to comply can result in legal penalties and loss of customer trust. Map your data flows explicitly to meet these jurisdictional requirements.
Read the full guide: Data Residency Laws: How to Keep Data in the Right Jurisdiction
What to do and how to stay safe: You.com
- Verify that the zero data retention policy meets your organization's privacy requirements before integration.
- Test the API endpoint with sample data to ensure compatibility with your existing SDK code.
- Review the service level agreement and failover mechanisms to understand potential downtime scenarios.
- Confirm that the US hosting requirement aligns with your data residency and compliance obligations.
Step-by-step guide: Secure Home Routers for Small Business Networks
General security guidance from the TechSec Headlines newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What models are available in You.com Inference Router?
The service routes requests across twelve live models, including GPT-OSS 120B, Kimi K3 and Qwen3.8 Max.
Does You.com store user data?
No, the service promises zero data retention, processing prompts and outputs in memory without storage.
Is the service hosted in the United States?
Yes, all processing occurs within the United States, with failover to vetted US-hosted partners.
