Pricing & Billing
Transparent, flexible pricing for MiroMind
Billing Model Overview
MiroMind uses a hybrid billing model combining Resource Packs (prepaid credits) and Pay-As-You-Go (consumption-based) charging.
Prepaid API call credits suitable for most standard API services.
Applicable to: MiroMind OS standard API interfaces
New users: Receive free trial credits upon registration
When your resource pack is exhausted or when using advanced services, the system automatically switches to pay-as-you-go mode, deducting costs directly from your account balance.
Pricing Structure
Our pricing is based on actual resource consumption, providing high cost-performance services. Costs consist primarily of Model Inference fees and Tool Usage fees.
| Model | Input | Output |
|---|---|---|
mirothinker-1-7-deepresearch-mini Mid-tier · 30B agent · cost-effective | $1.25 / 1M | $10 / 1M |
mirothinker-1-7-deepresearch Flagship · 235B agent · top-quality research | $4 / 1M | $25 / 1M |
Token counts include all internal agent activity — reasoning, tool result feedback, and summarization. See the Streaming Guide for the full activity breakdown.
| Tool | Public price |
|---|---|
Web search Charged each time the agent fetches a page (the fetch includes search → scrape → in-tool extraction). | $0.05 / fetch |
Pure URL discovery (Serper-style search that returns links without fetching) is free.
Sandbox execution, command-line, and other internal tools are bundled into the per-token rate above — no separate per-call charge.
Settlement & Charging Logic
To ensure service stability and fairness, we use a "pre-freeze + actual settlement" mechanism.
Resource Pack Check
System first checks if you have available resource pack credits. If sufficient, credits are deducted first.
Balance Pre-freeze
If pay-as-you-go is needed, the system pre-freezes an estimated amount before the request begins. Requests are rejected if balance is insufficient.
Actual Settlement
After task completion, the system calculates final costs based on actual tokens and tool calls. Unfreezes pre-charged amount and deducts actual cost (with refund for overcharge).
Task Failed
If a request fails due to platform or service issues, no charges are incurred and pre-frozen funds are fully refunded.
User Cancellation
If you cancel a task mid-execution, only the actual compute and tool costs incurred up to the cancellation point are charged.
Console Display
View the final total cost for each API call in your user console.
Audit Records
While the frontend shows only totals, our backend precisely records every token consumption and tool call, ensuring every charge is traceable and auditable.