Your AI Drive credits are the key to unlocking advanced capabilities within our platform. You can use your credits for a variety of tasks, including OCR (Optical Character Recognition) and processing with our AI models.
Each page you process with OCR consumes one credit.
For AI-powered tasks, credit use varies with the model, the amount of content, and the type of work. The table below provides a broad relative comparison of the standard token rates for the models currently available to you; it is not a credit estimate for a page or task.
Agent mode lets AI Drive plan and complete multi-step work across your documents and connected tools. Credit use reflects the work completed—selected model and reasoning level, pages or content read, tool calls, and any parallel or retried steps. Higher-capability or higher-reasoning choices can use more credits. Available models depend on your plan and your team's model-access settings.
- Model reasoning: Time spent by the selected model(s) thinking, planning, and generating results. Higher-capability models or higher reasoning settings can use more credits per unit of work than efficient models.
- Tools used: Each tool the Agent invokes (document reader, web fetch, data extraction, semantic search, comparison, etc.) contributes to usage based on its run time and complexity.
- Pages read: Credits scale with volume. Reading more pages across more files costs more than reading a few pages in a single file.
- Web/content retrieval: When the Agent fetches external pages or downloads files for analysis, those actions and any subsequent parsing/reading are counted.
- Parallel subtasks and retries: When the Agent branches work or retries a step to improve quality, those additional actions are included.
Chat Mode
Compare the models currently available to you. Relative model rate is a broad comparison of each model's standard token rate—not an estimate for a particular task.
| Chat Model | Model tier | Relative model rate |
| GPT-5 mini | Fast and efficient | |
| GPT-5.6 Luna | Fast and efficient | |
| Gemini 3 Flash Preview | Fast and efficient | |
| Claude 4.5 Haiku | Fast and efficient | |
| Claude Sonnet 5 | Powerful | |
| Gemini 3.1 Pro Preview | Powerful | |
| GPT-4.1 | Fast and efficient | |
| GPT-5.6 Terra | Powerful | |
| Kimi K3 | Powerful | |
| Claude Fable 5.1 | Powerful | |
| Claude Opus 5 | Powerful | |
| GPT-5.6 Sol | Powerful | |
| GPT-6 Astra | Powerful |
Agent Mode Credit Usage
Compare the Agent models currently available to you. Powerful models are best for tougher reasoning and multi-step analysis; fast, efficient models suit routine or high-volume work. Agent credits are based on actual tokens, reasoning, and tool work—not a flat per-page rate.
| Agent Model | Model tier | Relative model rate |
| GPT-6 Astra | Powerful | |
| Claude Fable 5.1 | Powerful | |
| Claude Opus 5 | Powerful | |
| GPT-5.6 Sol | Powerful | |
| Grok 4.6 | Powerful | |
| GPT-5.6 Terra | Powerful | |
| Claude Sonnet 5 | Powerful | |
| Gemini 3.1 Pro Preview | Powerful | |
| Kimi K3 | Powerful | |
| Gemini 3.8 Flash | Fast and efficient | |
| Claude 4.5 Haiku | Fast and efficient | |
| Gemini 3 Flash Preview | Fast and efficient | |
| GPT-5.6 Luna | Fast and efficient |