Skip to content
CheapestInference

MiMo V2.6 Flash API — unlimited & flat-rate access

MiMo V2.6 Flash is Xiaomi’s new-generation fast model — a 309B-parameter mixture-of-experts model with 15B active parameters, image input on the same endpoint and reasoning you switch on per request. CheapestInference serves it in the Core Pool through an OpenAI- and Anthropic-compatible API on unlimited, flat-rate time-block subscriptions — so your cost does not scale with tokens.

Quick facts

Section titled “Quick facts”
ModelMiMo V2.6 Flash
VersionV2.6-Flash — new generation, live since 2026-09-23, what changed
Model IDmimo-v2.6-flash
Context window1M tokens (1,048,576)
VisionYes — image input on the same endpoint
Max request size8 MB per request
Open weightsYes — XiaomiMiMo/MiMo-V2.6-Flash-RL, MIT license
Per-token price elsewhere$0.14 in / $0.28 out per 1M — reference only
PoolCore Pool (with DeepSeek V4.1 Flash)
Endpoints/v1/chat/completions (OpenAI), /anthropic/v1/messages (Anthropic)
Pricing$22/mo per 8-hour daily time block ($18.70/mo with annual billing), up to full 24/7

Call MiMo V2.6 Flash

Section titled “Call MiMo V2.6 Flash”
from openai import OpenAI


client = OpenAI(
    base_url="https://api.cheapestinference.com/v1",
    api_key="sk-..."  # your subscriber key
)


response = client.chat.completions.create(
    model="mimo-v2.6-flash",
    messages=[{"role": "user", "content": "Classify these support tickets..."}],
)
Terminal window
curl https://api.cheapestinference.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "mimo-v2.6-flash", "messages": [{"role": "user", "content": "Hello"}]}'

Image input uses the standard formats on the same endpoints: image_url content parts on /v1/chat/completions, or image blocks on /anthropic/v1/messages — in user messages, JPEG, PNG, GIF or WebP, within the 8 MB per-request limit. Audio and video input are not accepted on the API.

Reasoning is off by default — you get fast, direct answers. Enable it per request with "thinking": { "type": "enabled" } (both endpoints), or with the OpenAI reasoning_effort parameter (low, medium, high). With reasoning on, set a generous max_tokens so answers are not truncated mid-reasoning. JSON mode (response_format: { "type": "json_object" }) and tool calling work as on every other model.

Why flat-rate MiMo V2.6 Flash

Section titled “Why flat-rate MiMo V2.6 Flash”

MiMo V2.6 Flash keeps the size class and the list price of MiMo v2.5 while adding a generation of post-training: on Xiaomi’s own agent benchmarks it scores 87.6 on Terminal-Bench 2.1, 67.9 on DeepSWE v1.1 and 95.1 on CyberGym — within four points of the 1T-parameter MiMo-V2.6-Pro everywhere, and ahead of it on CyberGym (vendor-run numbers, not yet independently reproduced). That profile suits high-volume pipelines, extraction and classification, screenshots and diagrams alongside text, and agent loops that would be expensive on per-token billing. On CheapestInference it is billed at a flat monthly rate, not per token, so bulk workloads have a fixed, predictable cost — a Core Pool time block covers unlimited MiMo V2.6 Flash usage (plus DeepSeek V4.1 Flash) during your reserved hours. It works in any OpenAI-compatible client.

Common questions

Section titled “Common questions”

Is there a MiMo V2.6 Flash API? Yes. Use model id mimo-v2.6-flash against https://api.cheapestinference.com/v1. The API is OpenAI- and Anthropic-SDK compatible, and a Core Pool subscription gives unlimited usage — no token counting, no overage charges during your reserved daily 8-hour blocks — for a flat $22/month per block ($18.70/mo with annual billing).

What changed from MiMo v2.5? The Core Pool’s MiMo slot was upgraded in place to V2.6 Flash on 2026-09-23. It is a new generation at the same size class (309B total / 15B active) and the same list price: Xiaomi credits scaled reinforcement learning — about 750,000 trajectories across more than 7,000 environments — for a large jump on agent benchmarks (on DeepSWE v1.1, 67.9 where the previous generation’s Pro model scored 19.0). It also takes image input on the API. Same key, same endpoints, same subscription — every Core subscriber has it now. The full analysis is in MiMo-V2.6 Pro & Flash: specs, benchmarks, pricing.

Is mimo-v2.5 still accepted? Yes, for a grace period. Requests with the old id mimo-v2.5 are served by MiMo V2.6 Flash until 2026-10-23; after that date the id returns an invalid-model error. Migration is one line — change the model id to mimo-v2.6-flash. See MiMo v2.5 — retired.

How much does MiMo V2.6 Flash cost? From $22/month ($18.70/mo with annual billing). You reserve one or more 8-hour daily time blocks (up to full 24/7) and use MiMo V2.6 Flash with no token caps — billed at a flat monthly fee, not per token.

Can MiMo V2.6 Flash read images? Yes. Send image_url content parts (OpenAI format) or image blocks (Anthropic format) in user messages: JPEG, PNG, GIF and WebP, inside the same 8 MB request budget as everything else. Audio and video are not accepted on the API.

Does MiMo V2.6 Flash reason? On request. Reasoning is off by default for fast answers; add "thinking": { "type": "enabled" } or reasoning_effort (low, medium, high) to a request to turn it on, and read it back as reasoning_content (OpenAI endpoint) or a thinking block (Anthropic endpoint). Details on the Reasoning page.

How large can a request be? Each request to the Core Pool accepts up to 8 MB of payload — messages, history and attachments combined. Larger requests are rejected with a context_length_exceeded error; agent clients such as Claude Code compact the conversation automatically and retry. See Request size limits.

Is MiMo V2.6 Flash the same as MiMo-V2-Flash? No. MiMo-V2-Flash is Xiaomi’s late-2025 model with a 256K context. MiMo V2.6 Flash is the September 2026 successor to MiMo v2.5.

Who makes MiMo V2.6 Flash? MiMo V2.6 Flash is made by Xiaomi’s MiMo team and published under the MIT license on Hugging Face. CheapestInference serves it on flat-rate, unlimited time-block subscriptions in the Core Pool.

Is MiMo V2.6 Flash OpenAI-compatible? Yes. Use model id mimo-v2.6-flash against https://api.cheapestinference.com/v1 with the OpenAI SDK, or the /anthropic endpoint with the Anthropic SDK.

Section titled “Related”