Synthetic.new
by Synthetic
Run open-source LLMs privately in secure datacenters (US & EU); we never train on your data.
Features
- Privacy-first LLM hosting: runs open-source language models in secure datacenters located in the US and EU.
- Data non-training guarantee: the company states it does not train on customer data.
- Automatic API-data deletion policy: API customer data is auto-deleted within 14 days (per public policy snippets).
- API access for inference (details / endpoints not public in the sources found).
- Formal legal and privacy documents available (Terms of Service, Privacy Policy) on the website.
Superpowers
Synthetic.new is positioned for teams that need LLM inference without exposing sensitive data to large public-model vendors. Key advantages:
- Better data control and compliance posture for regulated data (PII, healthcare, finance) because models run in dedicated datacenters and the provider asserts no training on customer data.
- Use of open-source models can reduce vendor lock-in and potentially lower cost compared to proprietary hosted LLMs.
- Simple API-based integration model (inferred from marketing snippets) for embedding private LLMs into products and workflows.
Who it’s for
- Enterprises and startups that require private inference for sensitive data.
- Security- and privacy-conscious engineering teams that want an alternative to major public LLM APIs.
Pricing
Public pricing information was not available in the sources reviewed. The website appears to be the primary source for up-to-date pricing; likely options:
- Free / trial tier for evaluation (unknown)
- Usage-based pricing (inference tokens / compute hours)
- Dedicated / enterprise plans with SLA for datacenter regions (US / EU)
Recommendation: contact sales or request pricing via the site or docs for exact tiers and committed-use discounts.
API / Documentation
- The site (https://synthetic.new) redirects to a landing page and exposes Terms/Privacy pages but I could not find a public, detailed API reference or SDK links in the materials available during this research pass.
- If you need to evaluate integration: look for docs or developer portal links on the landing page, or sign up for an account to access API keys and example requests.
Practical usage examples
- Private customer-support assistant: route customer transcripts through a private LLM hosted in your chosen datacenter region to avoid sending transcripts to large public model providers.
- PII-safe data augmentation: generate synthetic variations of data for ML testing while ensuring raw records are not retained beyond the stated retention window.
- Regulated-model inference: run ML-powered classification or extraction on health/finance data where data residency (EU vs US) and retention controls matter.
What I couldn’t find / verification needed
- Public, detailed API documentation or SDK examples.
- Clear, public pricing and tier definitions.
- A full list of supported models (names / sizes) available for inference.
- SLA, throughput, and latency guarantees for enterprise customers.
- Company background / leadership / case studies.
Suggested next steps
- Visit the landing page and developer/docs sections at https://synthetic.new and sign up for an evaluation account to access API docs.
- Review the Terms of Service and Privacy Policy for contractual guarantees about data handling.
- Request a sales demo / security datasheet for SOC/ISO certifications and data residency details.
- Run a short POC to validate latency, supported models, and data deletion behavior.