AI providers and how they handle your data
Choose a model with the price and data handling that suit you. Choosing a model means accepting its provider’s terms and data policy.
- Zero data retention
- The model provider does not retain the inputs or outputs of your request or use them to train models.
- No training
- Your inputs and outputs are not used to train models, but the provider may retain them for a limited time.
- May keep and train
- The provider may keep your inputs and outputs and may use them to train models.
ZDR is the most preferable level. These levels describe the model provider. They do not change how this site saves your conversations; see our Privacy Policy.
Conversation titles: when you start a saved conversation with AI questions or your agent, its first message also goes to DeepSeek V4 Flash (OpenRouter, Zero data retention), whichever model you chose, to write the short title in your list. Nothing else from the conversation is sent for the title, and it costs a small fraction of a cent from your AI spending.
OpenAI
Checked on .
| Data level | No training | ||
|---|---|---|---|
| Training | API inputs and outputs are not used for training unless the customer opts in; Focus on Foundations has not opted in. | ||
| Retention | Abuse-monitoring logs may be retained for up to 30 days. We disable saved API responses for answers and agent steps. | ||
| ZDR status | Requested, not granted | ||
| Policies | Data policy · Terms | ||
| Model | Input / million tokens | Output / million tokens | Data level |
| GPT-6 Astragpt-6-astra | $10 | $50 | No trainingRetention: Abuse-monitoring logs may be retained for up to 30 days. We disable saved API responses for answers and agent steps.Training: API inputs and outputs are not used for training unless the customer opts in; Focus on Foundations has not opted in.Checked on . Source 1 |
| GPT-6 Solgpt-6-sol | $2 | $10 | No trainingRetention: Abuse-monitoring logs may be retained for up to 30 days. We disable saved API responses for answers and agent steps.Training: API inputs and outputs are not used for training unless the customer opts in; Focus on Foundations has not opted in.Checked on . Source 1 |
| GPT-6 Lunagpt-6-luna | $0.1 | $0.5 | No trainingRetention: Abuse-monitoring logs may be retained for up to 30 days. We disable saved API responses for answers and agent steps.Training: API inputs and outputs are not used for training unless the customer opts in; Focus on Foundations has not opted in.Checked on . Source 1 |
| GPT-5.4 Progpt-5.4-pro | $30 | $180 | No trainingRetention: Abuse-monitoring logs may be retained for up to 30 days. We disable saved API responses for answers and agent steps.Training: API inputs and outputs are not used for training unless the customer opts in; Focus on Foundations has not opted in.Checked on . Source 1 |
| GPT-4ogpt-4o | $2.5 | $10 | No trainingRetention: Abuse-monitoring logs may be retained for up to 30 days. We disable saved API responses for answers and agent steps.Training: API inputs and outputs are not used for training unless the customer opts in; Focus on Foundations has not opted in.Checked on . Source 1 |
| GPT-4o minigpt-4o-mini | $0.15 | $0.6 | No trainingRetention: Abuse-monitoring logs may be retained for up to 30 days. We disable saved API responses for answers and agent steps.Training: API inputs and outputs are not used for training unless the customer opts in; Focus on Foundations has not opted in.Checked on . Source 1 |
| o3 minio3-mini | $1.1 | $4.4 | No trainingRetention: Abuse-monitoring logs may be retained for up to 30 days. We disable saved API responses for answers and agent steps.Training: API inputs and outputs are not used for training unless the customer opts in; Focus on Foundations has not opted in.Checked on . Source 1 |
Anthropic
Checked on .
| Data level | No training | ||
|---|---|---|---|
| Training | API inputs and outputs are not used for training without express permission; Focus on Foundations has not given that permission. | ||
| Retention | Inputs and outputs are deleted within 30 days by default. Neither Claude Sonnet 5.5 nor Claude Opus 5.5 is a Covered Model requiring 30-day retention. | ||
| ZDR status | Available only by arrangement; not in place | ||
| Policies | Data policy · Training policy · Terms | ||
| Model | Input / million tokens | Output / million tokens | Data level |
| Claude Sonnet 5.5claude-sonnet-5-5 | $2 | $10 | No trainingRetention: Inputs and outputs are deleted within 30 days by default. Neither Claude Sonnet 5.5 nor Claude Opus 5.5 is a Covered Model requiring 30-day retention.Training: API inputs and outputs are not used for training without express permission; Focus on Foundations has not given that permission.Checked on . Source 1 |
| Claude Opus 5.5claude-opus-5-5 | $4 | $20 | No trainingRetention: Inputs and outputs are deleted within 30 days by default. Neither Claude Sonnet 5.5 nor Claude Opus 5.5 is a Covered Model requiring 30-day retention.Training: API inputs and outputs are not used for training without express permission; Focus on Foundations has not given that permission.Checked on . Source 1 |
OpenRouter
Checked on .
| Data level | Per model, enforced on every request | ||
|---|---|---|---|
| Training | Training policy depends on the model and its eligible providers; see each model below. | ||
| Retention | Retention depends on the model. Each request uses the routing restriction that matches its disclosed data level. | ||
| ZDR status | Enforced per request for models marked ZDR | ||
| Policies | Data policy · Terms | ||
| Model | Input / million tokens | Output / million tokens | Data level |
| DeepSeek V4 Flashdeepseek/deepseek-v4-flash | $0.14 | $0.28 | ZDRRetention: The routed provider does not retain your inputs or outputs.Training: The routed provider does not use your inputs or outputs to train models.Checked on . Source 1 Source 2 |
| Kimi K2.6moonshotai/kimi-k2.6 | $0.95 | $4 | ZDRRetention: The routed provider does not retain your inputs or outputs.Training: The routed provider does not use your inputs or outputs to train models.Checked on . Source 1 Source 2 |
| GLM 5.2z-ai/glm-5.2 | $0.6496 | $2.0416 | ZDRRetention: The routed provider does not retain your inputs or outputs.Training: The routed provider does not use your inputs or outputs to train models.Checked on . Source 1 Source 2 |
| Nemotron 3 Supernvidia/nemotron-3-super-120b-a12b | $0.08 | $0.45 | ZDRRetention: The routed provider does not retain your inputs or outputs.Training: The routed provider does not use your inputs or outputs to train models.Checked on . Source 1 Source 2 |
| Qwen 3.7 Plusqwen/qwen3.7-plus | $0.32 | $1.28 | No trainingRetention: Not stated by the providerTraining: Alibaba does not use prompts to train models.Checked on . Source 1 Source 2 |
| Qwen 3.7 Maxqwen/qwen3.7-max | $1.475 | $4.425 | No trainingRetention: Not stated by the providerTraining: Alibaba does not use prompts to train models.Checked on . Source 1 Source 2 |
Inworld
Checked on .
Used only on the owner’s computer to record lesson voices ahead of time. It is not a model a person chooses on the site.
| Data level | Not used for site requests | ||
|---|---|---|---|
| Training | No account data is sent. | ||
| Retention | No account data is sent. | ||
| ZDR status | Not applicable to site requests | ||
| Policies | Data policy · Terms | ||
Model availability depends on your account’s access. Prices are per million tokens; actual OpenRouter route prices and long-context charges may vary.