SAME PRICE AS V2.5MIT WEIGHTS309B / 15B ACTIVETEXT, IMAGE, VIDEO, AUDIO IN1M CONTEXT
MiMo-V2.6-Flash API Pricing
The high-volume half of Xiaomi's V2.6 release costs exactly what MiMo-V2.5 did: $0.14/M input on a cache miss, $0.0028/M on a hit and $0.28/M output, with a new Batch API at half. It reads text, images, video and audio into a 1M-token window, and its weights are public under the MIT licence: 309B parameters, 15B active per token. Xiaomi says it beats V2.5-Pro across the board, a model that lists at 3.1x its rates. V2.5 was retired on October 21, 2026, after a week of Xiaomi routing its requests here.
Input - per 1M tokens
$0.14/M
Batch $0.07/M cache miss
Output - per 1M tokens
$0.28/M
Batch $0.14/M
Cached input - per 1M tokens
$0.0028/M
Cache writes free for now 98% off
Effective - agentic blend
$0.0477/M
92/8 split - 82% cache
§ 01 / TERMINAL
MiMo-V2.6-Flash API cost calculator.
Calculator pre-loaded with Xiaomi's overseas USD list for MiMo-V2.6-Flash. The Batch API halves every line for work that can wait, and cache writes are currently free.
$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
§ 02 / SCENARIOS
MiMo-V2.6-Flash cost on real workloads.
AGENT LOOP
Repo-wide bug fix
$0.0061/task
VIDEO + AUDIO
Meeting recording summary
$0.00249/meeting
RAG SUPPORT
Support agent ticket triage
$0.0005/ticket
CLASSIFY
Catalogue item tagging
$0.00011/item
Estimate · xiaomi-tokenizer-estimate · ≈3.85 chars/token Auto-counts as you type
This is a chars-per-token approximation, not a real tokenizer. Actual tokens vary by language, code density, and tool-call overhead — counts are typically ±10–20% off for English prose, more for code or non-Latin scripts. For exact billing, use the vendor's official tokenizer.
Characters 521
Words 86
Tokens (estimated) 135 tokens
Cost as input · uncached $0.00002 USD
Cost as output · uncached $0.00004 USD
Cost as cached input $0.0000004 USD
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| MiMo-V2.6-Flash Current | $0.14 cache $0.0028 | $0.28 | $0.0477 agentic 92/8 | 1M | Open-weight omni-modal model for high-volume calls |
| MiMo-V2.6-Pro | $0.435 cache $0.0036 | $0.87 | $0.144 +3.0× | 1M | The flagship, for long and hard agent runs |
| MiMo-V2.5 | $0.14 cache $0.0028 | $0.28 | $0.0477 retired | 1M | What this model replaces, for the record |
| DeepSeek V4.1 Flash | $0.30 cache $0.006 | $1.20 | $0.15 +3.1× | 1M | Cheap 1M-context agent loops with images |
| GPT-6 Luna | $0.10 cache $0.01 | $0.50 | $0.0682 +43% | 1.05M | High-volume work on a closed frontier stack |
| Mistral Small 4 | $0.15 cache $0.015 | $0.60 | $0.0842 +76% | 256K | Low-cost general work from a European vendor |
| Gemini 3.8 Flash | $0.75 cache $0.075 | $3.75 | $0.481 +10× · promo | 1M | Multimodal input from a hyperscaler stack |
MiMo-V2.6-Flash has published scores on 13 benchmarks we track, and leads CyberGym. All of them are the lab's own reports; each row links to the page that published it.
| Benchmark | Score | Rank | Leader | Source |
|---|---|---|---|---|
| CyberGym Cybersecurity | 95.1% as reported | #1 of 8 | MiMo-V2.6-Flash leads | Vendor-reported Xiaomi |
| AutomationBench Agents and computer use · Main set (as reported) | 52.3% v1.0.6 | #5 of 22 | MiniMax M3.1 Flash · 62.8% | Vendor-reported Xiaomi |
| JobBench Professional work | 61.2% as reported | #2 of 5 | MiMo-V2.6-Pro · 62% | Vendor-reported Xiaomi |
| Terminal-Bench 2 Agents and computer use · Terminal-Bench 2.1 | 87.6% as reported | #7 of 17 | DeepSeek V4.1 Flash · 90.6% | Vendor-reported Xiaomi |
| DeepSWE Coding | 67.9% as reported | #12 of 27 | Gemini 4 Argon · 77.9% | Vendor-reported Xiaomi |
| Toolathlon Agents and computer use | 73.6% as reported | #4 of 9 | MiMo-V2.6-Pro · 76.9% | Vendor-reported Xiaomi |
| OSWorld-Verified Agents and computer use | 80.8% as reported | #4 of 8 | Kimi K3 · 84.8% | Vendor-reported Xiaomi |
| ProgramBench Coding | 26% as reported | #4 of 7 | Kimi K3.1 Pro · 31.5% | Vendor-reported Xiaomi |
| Terminal-Bench 4.0 Agents and computer use | 28.8% as reported | #20 of 34 | Kimi K3.1 Pro · 65.7% | Vendor-reported Xiaomi |
| ExploitGym Cybersecurity | 6% as reported | #7 of 9 | Gemini 4 Carbon · 54.8% | Vendor-reported Xiaomi |
| ExploitBench Cybersecurity | 25.3% as reported | #6 of 7 | GPT-6 Astra · 100% | Vendor-reported Xiaomi |
| SEC-bench Pro Cybersecurity | 47.5% as reported | #7 of 8 | Gemini 4 Carbon · 88.2% | Vendor-reported Xiaomi |
| Agents' Last Exam Agents and computer use · Split not stated | 27.6% split not stated | #4 of 4 | MiMo-V2.6-Pro · 31.6% | Vendor-reported Xiaomi |
MiMo-V2.6-Flash API pricing FAQ.
What MiMo-V2.6-Flash costs, how it relates to V2.5 and V2.6-Pro, and what V2.5's retirement on October 21 means for code that still names the old id.
Q · 01 How much does MiMo-V2.6-Flash cost? +
On Xiaomi's overseas list it is $0.14/M input on a cache miss, $0.0028/M on a cache hit and $0.28/M output. The Batch API halves all three:
$0.07 / $0.0014 / $0.14. Under our 92/8 agentic blend with 82% cache hits the planning figure is $0.05/M. The dollar list and Xiaomi's separate yuan list for China are two price lists, not one converted into the other.Q · 02 Is it more expensive than MiMo-V2.5? +
No. The pricing table prints
mimo-v2.6-flash and mimo-v2.5 on one row with one set of rates. The release adds a Batch API at half price and MIT-licensed weights, and keeps the 1M window and the text, image, video and audio inputs.Q · 03 What happens to MiMo-V2.5? +
Requests that name
mimo-v2.5 fail since 10:00 Beijing time on October 21, 2026 (02:00 UTC), when Xiaomi retired the name. In the week before, Xiaomi routed them here automatically - from 18:00 Beijing time on October 14 they were served and billed as mimo-v2.6-flash, at the same rates, with this model's defaults where a request set none (a 131,072-token default for max_completion_tokens against V2.5's 32,768). Xiaomi had first said there would be no system replacement and changed that on October 9. Code that still names the old id now needs the new name.Q · 04 Flash or Pro? +
MiMo-V2.6-Pro lists at about 3.1x Flash on input and output. Xiaomi's launch post says Flash comprehensively outperformed the previous generation's Pro, and puts V2.6-Pro level with frontier closed models on most agent benchmarks; those are the vendor's claims, not our measurements. Flash is the default for high-volume calls, classification, support and media summarisation; Pro is for long-horizon agent runs where a failed attempt costs more than the tokens.
Q · 05 Are the weights open, and how big is the model? +
Yes. The weights are on Hugging Face as
XiaomiMiMo/MiMo-V2.6-Flash-RL under the MIT licence. The model card describes a sparse mixture-of-experts with 309B total and 15B active parameters (256 routed experts, 8 active), a 681M-parameter vision encoder and audio encoders, and a 1M-token context.Q · 06 What does it accept, and what are the limits? +
Text, images, video and audio in; text out. The context window is 1M tokens with up to 128K output tokens, and pay-as-you-go accounts get 100 requests per minute and 10M tokens per minute. Thinking mode, tool calls, structured output, streaming and context caching are supported. Web search is billed separately at $5 per 1,000 calls overseas.
Q · 07 What do cache writes cost? +
Nothing for now: Xiaomi lists cache writes as limited-time free, with no end date and no future price, so this page does not model one. A cache hit costs
$0.0028 per million, 2% of a miss, so on repeated context this model is cheaper than its input rate suggests.