MiMo-V2.5
Xiaomi’s open model that reads, looks and listens — text, images, audio and video.
About MiMo-V2.5
MiMo-V2.5 is one of the few open-weight models that understands audio and video as well as text and images, using encoders Xiaomi built itself. It is a 310-billion-parameter mixture of experts with 15 billion active per token and a million-token context, released under MIT in April 2026.
Xiaomi reports 56.1% on SWE-bench Pro and 65.8% on Terminal-Bench 2, respectable for its cost. Its weak spot is putting those senses to work: on agent tasks that need it to act on images or video, it scores far lower.
Where it’s strong
Analysing recordings, video and images together with text.
Coding agents on a budget.
Media and document analysis.
Where to be careful
Weak at multimodal agent tasks: 23.8% on Xiaomi’s own Claw-Eval Multimodal.
Xiaomi has not said how its reasoning can be controlled.
Its scores are Xiaomi’s own.
Price
For scale: 1,000 requests of about 3,000 tokens in and 1,000 out cost $0.91 — a long prompt and a page of answer each. Every rate is published and stays the same from one request to the next. See every model.
Specs
mimo-v2.5Call it
Use the OpenAI SDK you already have — point it at Roar AI and name the model. Read the docs.
from openai import OpenAI
client = OpenAI(base_url="https://api.roar-ai.com/v1", api_key="roar_live_…")
reply = client.chat.completions.create(
model="mimo-v2.5",
messages=[{"role": "user", "content": "Summarise this contract in plain English."}],
)
Questions
How much does MiMo-V2.5 cost?
Input is $0.182 per 1M tokens; output is $0.364 per 1M tokens. For scale, 1,000 requests of about 3,000 tokens in and 1,000 out cost $0.91. The rate is published and does not change from one request to the next.
Can I use MiMo-V2.5 alongside other models?
Yes. Every model on Roar AI is on the same API key and the same monthly invoice, so switching from MiMo-V2.5 to another model is a change to one word in your code.
Is MiMo-V2.5 open source?
Its weights are open under the MIT licence, so you could run it yourself. Through Roar AI you call it like any other model, with no servers to manage.