Credits & Usage
Understand the credit system — how credits are calculated, consumed, tracked, and what limits apply.
What Are Credits?
Credits are the unified currency on Assistant Core. Every AI operation — chat, voice, realtime — consumes credits from your wallet. More expensive models cost more credits, cheaper models cost fewer — you only pay for what you actually use.
- Pay for what you use — Premium models cost more credits, efficient models cost fewer
- One wallet for everything — Chat, voice, and realtime all share the same credits
- Flexible monthly budget — Use your credits however you want throughout the month
Credits Per Plan
Each plan includes a different monthly credit allowance. See Plans & Pricing for the latest credits, feature access, and resource limits.
Credits are user-global. All AI operations across all assistants deduct from the same credit wallet.
How Are Credits Calculated?
Each AI operation is charged credits based on the model and usage:
Chat (LLM)
Credits are calculated by token count (input and output):
| Factor | Description |
|---|---|
| Input tokens | Your message + conversation context |
| Output tokens | The AI's response |
| Cached tokens | Tokens from previously processed context |
Example: A typical chat message (~500 input + 300 output tokens) with DeepSeek V4 Flash costs about 1 credit.
Voice
| Type | How It's Charged |
|---|---|
| Speech to Text (ASR) | By audio duration (minutes or hours) |
| Text to Speech (TTS) | By character count or audio duration |
| Realtime Voice | By session duration (minutes) |
Utility & Embedding
Some background features also consume credits in very small amounts:
| Feature | When |
|---|---|
| Follow-up suggestions | After each chat turn (if enabled) |
| Title generation | First message of each conversation |
| Long Memory | Extracting key information from conversations |
| Knowledge Base search | Finding relevant content in your documents |
| Memory recall | Recalling previously saved information |
Utility and embedding costs are very small (typically under 1 credit per operation) and are calculated automatically — you don't need to worry about them.
Image Generation
| Model | Quality | Credits / image (estimate) |
|---|---|---|
| GPT Image Mini | Auto / Low | ~8 |
| GPT Image Mini | Medium | ~23 |
| GPT Image Mini | High | ~54 |
| Gemini Flash Image | Default | ~59 |
Web Search
| Type | Credits / query |
|---|---|
| Search (web, map, video, scholar) | ~2 |
| Webpage scrape | ~3 |
Estimated Credits Per Model
| Model | Credits / message (estimate) |
|---|---|
| DeepSeek V4 Flash | ~1 |
| MiMo V2.5 | ~1 |
| GPT-4.1 Mini | ~2–5 |
| GPT-5.6 Luna | ~5–15 |
| Gemini 3.5 Flash | ~5–12 |
| GPT-4.1 | ~5–15 |
| Grok 4.5 | ~8–20 |
| Claude Sonnet 4.6 | ~10–25 |
| GPT-5.6 Terra | ~15–40 |
These are estimates. Longer conversations build up more context, which increases credit cost per message. Tool calls also add to the total.
Resource Limits
In addition to credits, each plan has resource limits such as Knowledge Base documents, devices, users per assistant, and owned assistants. These limits are based on the assistant owner's plan, not the user who is chatting.
See Plans & Pricing for the latest resource limits.
Tracking Your Credits
- Go to My Subscription
- The Credits bar shows:
- Used — Credits consumed in the current period
- Remaining — Available credits
- Limit — Total credits granted for this period
When Limits Are Reached
Credits exhausted
- Chat — Cannot send new messages
- Voice — Cannot start new voice sessions
- Realtime — Realtime voice sessions are rejected
Resource limits reached
- KB Documents — Can't upload more files until you delete some or upgrade
- Devices — Can't register new devices
- Assistants — Can't create new assistants until you delete some or upgrade
Zero credits = AI usage paused. Upgrade your plan for more credits, or wait for the next billing period to receive a fresh allocation.
Tips to Save Credits
- Choose efficient models — DeepSeek V4 Flash and MiMo V2.5 deliver great results at the lowest cost (~1 credit/message)
- Disable suggestions if you don't need them — saves ~1 credit per chat turn
- Use Knowledge Base instead of pasting long content into chat — reduces input tokens
- Use Realtime Voice only when you need the real-time experience — Standard Voice is usually cheaper
Need More Credits?
Compare plans to find the right fit, or contact us for an Enterprise plan with unlimited credits.