MiMo 2.6 Pricing: Choose Between Pro, Flash and UltraSpeed

MiMo 2.6 pricing depends on the exact model and billing route. Standard Pro, Flash and Pro UltraSpeed are not interchangeable: UltraSpeed’s published token rates are ten times standard Pro’s rates. Check the full identifier before deciding that a screenshot shows a sudden price increase.

The tables below reproduce the official pay-as-you-go schedules checked on September 22, 2026. They describe Xiaomi’s direct service, not Ofox pricing, a gateway discount or a desktop membership fee.

Overseas API rates in US dollars

Source: Xiaomi MiMo API pricing. UltraSpeed has separate access arrangements; a price listing alone does not establish that it is enabled on your account.

China API rates in yuan

These are the provider’s published currency schedules, not exchange-rate conversions made for this article. Use the schedule and region that actually apply to your account. Do not mix a yuan input price with a dollar output price in a cost comparison.

What a cache hit changes

Consider a hypothetical standard Pro request on the overseas schedule: 20,000 uncached input tokens, 80,000 cached input tokens and 5,000 billed output tokens.

Uncached input: 20,000 / 1,000,000 × $0.435 = $0.00870
Cached input:   80,000 / 1,000,000 × $0.0036 = $0.000288
Output:          5,000 / 1,000,000 × $0.87  = $0.00435
Token subtotal                                      $0.013338

This is a calculation example, not an observed bill. With no cache hits and the same total input, the token subtotal would be $0.04785. Actual requests must use reported usage; merely repeating a prompt is not a guarantee of a cache hit.

If cached input is included within a total-input field, subtract it before calculating uncached charges. Also include billed reasoning, not just the answer visible in the chat interface. Tool fees and retries can make the completed-task bill larger than this single-request example.

Batch, search and Token Plan

The official Batch guide lists a 50% rate for supported Pro and Flash batch processing. UltraSpeed is not included. Batch uses the cash-balance API path rather than consuming Token Plan credits; use the base URL provided for your account.

The pricing page separately lists web-search charges and describes cache writing as temporarily free. These conditions have their own scope. Do not describe all tools as included in the token rate or convert a temporary condition into a permanent promise.

Token Plan and ordinary cash-balance billing are separate accounting systems. Before buying a plan, check the applicable key type, endpoint, included models and how usage is deducted. A large advertised credit count cannot be compared with token totals until the deduction rules are clear.

Which version should you evaluate first?

Choose a representative task with a measurable completion condition. Flash is the lower listed-price starting point; standard Pro is a separate candidate if the task needs a different level of reasoning. Evaluate UltraSpeed only when its access and latency benefit fit your use case. Higher throughput does not by itself establish lower cost per completed task.

For an agent, log every request, cached portion, billed output, retry and tool charge. Include the final acceptance result so a cheap failed attempt is not mistaken for an economical solution.

Use the MiMo access guide to separate OpenCode Free, Desktop and direct API routes. The API setup guide explains matching keys and endpoints before measuring usage.