Every model.
Nothing hidden.
One gateway to models from 8 companies, each with a plain YES/NO sheet on training, retention and where your data goes. Read it before you send a word.
Latest models
gpt-6.1-solOpenAI-9 Oct 2026GPT 6 Astragpt-6-astraOpenAI-9 Oct 2026Claude Haiku 4.5claude-haiku-4-5Anthropic-9 Oct 2026Ling 3.1 Flashling-3.1-flashInclusion AI-9 Oct 2026Muse Glimmer 30bmuse-glimmer-30bMetavia SaaufAI Cloud131K9 Oct 2026Jevjev-latestTypeSafe AI-9 Oct 2026GPT OSS 20bgpt-oss-20bOpenAIvia SaaufAI Cloud-9 Oct 2026GLM 5.3 Flashglm-5.3-flashZ.aivia SaaufAI Cloud1M8 Oct 2026Companies
Who builds the models in the catalog. Open one to see everything it has listed here.
How to read a disclosure
The same checklist for every model, so you can compare like with like. Squares, not stars: a Yes is not a score.
The statement applies to this model.
The statement does not apply.
Not confirmed yet. Assume the less private option.
- 01Are prompts and outputs stored by SaaufAI?Stored
- 02May the upstream provider use data for model training?Training
- 03Does the upstream provider retain data for abuse monitoring?Retention
- 04Is data processed outside the EU/EEA?Outside EU
- 05Is the model suitable for personal or sensitive data?Sensitive OK
- 06Does the model support image input?Images
- 07Does the model support tool / function calling?Tools
- 08Is the model output content-filtered by the provider?Filtered
- 09Does SaaufAI enforce its own safeguards on this model?Safeguards
- 10May requests be answered by a different (fallback) model?Fallback
What SaaufAI keeps
The gateway itself is deliberately boring with your data. Each provider's rules are in the model's disclosure sheet.
Never stored
- The text of your prompts and conversations
- The model's responses
- Images or files you send
- Anything sold or shared for advertising
Kept briefly, for operations
- Which model was used and when
- Token counts, latency and status codes
- Short error messages from providers
- Nothing that identifies the content
Questions, answered
Does SaaufAI store my prompts or the answers?
No. Content passes through in memory and is forwarded to the model provider. SaaufAI only keeps operational metadata such as timestamps, token counts and status codes. See the Privacy Policy.
What do the YES/NO disclosures mean?
Every model is checked against the same questions: training use, retention, data location, capabilities and more. A filled square means Yes, an empty one No. Neither is a rating; read the question. How to read disclosures.
What does “Not specified” mean?
The answer hasn't been confirmed yet. Treat it as unknown and assume the less private option until it is filled in.
Who actually runs the model?
Each model page names the company that made it and answers how your data is handled. The serving provider's own terms also apply to your requests.
How often is this information updated?
Disclosures are reviewed whenever a provider changes its terms or a model is added. Policies show their effective date at the top.