MiMo V2.5 Pro
Xiaomi’s flagship agentic reasoning model for complex coding and long-horizon tool use.
Quantization
FP8
Open weights
Context Window
1.0M
Tokens
Intelligence
43
AA Index
Released
Apr 2026
22 Apr 2026
Frontier reasoning, open weights
MiMo V2.5 Pro is Xiaomi’s flagship 1.02T-parameter MoE reasoning model for agentic workflows, complex software engineering, long-horizon tool use and long-context tasks.
Sparse activation keeps serving cost low, so a 1M-token context window and multi-step agent runs stay affordable across realtime, async and batch tiers.
1M context · 1.02T MoE
Best for repository-scale engineering
Complex Coding
Repository-scale engineering: multi-file edits, large refactors and iterative debugging across long sessions.
Agentic Workflows
Configurable reasoning effort for long-horizon planning, tool orchestration and recovery across many steps.
Tool Use
Enhanced structured generation for dependable function calling across long tool chains.
Long-Context Research
A 1M-token window keeps entire repositories, specs and document sets in a single request.
Also strong on long-context reasoning and enhanced structured generation.
Flexible Pricing Tiers
Choose the optimal balance of speed and cost for your workflow. Prices are per 1M tokens.
| Tier | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
| Batch | $0.22 | $0.44 |
| Async | $0.33 | $0.65 |
| Realtime | $0.44 | $0.87 |
Context window natively supported up to 1,048,576 tokens. We can offer custom pricing for bulk discounts, large workloads, and dedicated deployments — reach out to hello@doubleword.ai.
Start Building in Minutes
MiMo V2.5 Pro is accessible via OpenAI-compatible endpoints. Default sampling: temperature=1.0, top_p=0.95.
from openai import OpenAI
client = OpenAI(
api_key="your-api-key-here",
base_url="https://api.doubleword.ai/v1"
)
# Long-horizon agentic tool use
response = client.chat.completions.create(
model="XiaomiMiMo/MiMo-V2.5-Pro",
messages=[
{"role": "user", "content": "Plan and execute a multi-step refactor across this repository."}
],
temperature=1.0,
top_p=0.95,
)
print(response.choices[0].message.content)💡 Pro Tip
Route long-running agent steps through the async tier and bulk evals through batch — same model, same API, at half the realtime rate. See the docs for the batch job format.