MiMo-V2.6-Pro API pricing and cost calculator

By TechCompare · Updated · How we check

Xiaomi's direct half-cached 100K-input, 5K-output example costs $0.02628 per request, or $262.80 for 10,000 calls.

Calculator

Cost Comparison

Based on 100,000 input tokens (50% cached), 5,000 output tokens, and 100 requests.Prices come live from OpenRouter at each maker's own price where it's listed. Temporary deals are left out.

Loading model prices

MiMo-V2.6-Pro costs $0.435 per million uncached input tokens and $0.87 per million output tokens on Xiaomi's overseas pay-as-you-go API. Input that hits the prompt cache costs $0.0036 per million tokens. These are Xiaomi's published USD rates. The chart uses the same direct provider prices, so its starting estimate matches the calculations below.

For an agent or document workflow, the useful budget includes every call needed to finish the task, including reasoning output, tool responses, and retries.

Xiaomi: MiMo-V2.6-Pro

Cheapest host: GMICloud (bf16)

Input
$0.435
Output
$0.87
Cached
$0.0036

per 1M tokens

Price checked 2026-11-19

How this is calculated

A request with 100,000 input tokens and 5,000 output tokens costs $0.0435 for uncached input plus $0.00435 for output, totaling $0.04785. If half the input hits the cache, input comes to $0.02193 and the request total is $0.02628. At 100 monthly requests, the starting chart is $2.628, displayed as $2.63.

At 10,000 identical requests, that's $262.80 with half the input cached or $478.50 without cache hits. A fully cached version of that prompt costs $0.00471 per request, including the same output. This makes reusable context valuable, while generated output becomes a larger share of the total as the input bill falls. Xiaomi lists cache writes as free for a limited time.

That policy should be checked again before building a long-term budget around it. A cache-hit percentage describes input tokens actually reused, not a discount applied to the entire request. Output remains billed at the real-time rate.

Verdict

Evaluate Pro where complex tasks might justify spending more per request than Flash. Use the same prompts and success criteria, then record accepted results, output usage, repeated tool calls, and time spent reviewing them.

A lower failure rate can matter more than the price of one attempt, but that improvement needs to show up in your application. Stable prompt prefixes can keep input spending low. Keep generated explanations and repeated tool results within useful limits because cache savings don't remove the output bill.

More API Standalones scenarios

Frequently asked questions

What does MiMo-V2.6-Pro cost directly from Xiaomi?
The overseas real-time API charges $0.435 per million uncached input tokens, $0.0036 per million cache-hit tokens, and $0.87 per million output tokens. Apply each rate to its own usage category. These USD prices come from Xiaomi's pay-as-you-go table, rather than a converted domestic price or a subscription credit allowance.
How much does MiMo Pro save through prompt caching?
A cache hit costs about 99% less than uncached input. That percentage applies only to the tokens that hit the cache. With 100,000 input tokens, half cached, and 5,000 output tokens, the request falls from $0.04785 to $0.02628. Use the cache usage returned by the API to check whether your actual requests reach that assumption.
What if all the input is read from cache?
For 100,000 cache-hit tokens and 5,000 output tokens, Pro costs $0.00036 for input plus $0.00435 for output, totaling $0.00471. At 10,000 calls, that's $47.10. This is a real-time token example with confirmed cache hits. The initial request that creates a reusable prefix still needs to be counted using its actual cache-miss usage.
Are Xiaomi Token Plan credits the same as API pricing?
No. Xiaomi says ordinary pay-as-you-go API keys consume account balance separately from Token Plan quotas. The chart estimates the published overseas USD token charges. A monthly coding subscription needs its own calculation based on plan credits and eligible usage. Dividing a plan's fee by a token allowance can misrepresent a pay-as-you-go application bill.
What charges are outside this estimate?
The examples cover text input, cache hits, and output tokens. Xiaomi bills overseas web search separately at $5 per 1,000 uses, so 10,000 such uses add $50. Audio services have their own billing rules. Count additional model calls, output generated during reasoning, and unsuccessful attempts from your actual usage before turning a request estimate into a project budget.