MiMo-V2.5-Pro

Creator

Xiaomi

Released

2026-04-22

Intelligence

42.2

Artificial Analysis Index

Coding

60.2

Artificial Analysis Index

In $/1M

$0.43

input tokens

Out $/1M

$0.87

output tokens

Blended $/1M

$0.54

3:1 blended

Speed

63

tokens / sec

Profile

License, openness, and modality — the governance layer.

License

MIT

commercial OK

Weights

Open

downloadable

Modalities

Text

Parameters

1.0T

total

MiMo-V2.5-Pro is Xiaomi's 1.02T-parameter sparse Mixture-of-Experts language model with 42B active parameters and a 1M-token context window. It inherits the MiMo-V2-Flash hybrid-attention and Multi-Token Prediction design, extends context during pre-training up to 1M tokens, and uses supervised fine-tuning, domain-specialized reinforcement learning, and Multi-Teacher On-Policy Distillation to improve complex software engineering, long-horizon agentic tasks, and ultra-long-context coherence.

Capability profile

Category strength across 20 domains, via LLM Stats.

legal
100finance
100agents
100reasoning
100general
100language
90math
80healthcare
80frontend development
80
code
70biology
70physics
70chemistry
70tool calling
70long context
60coding
40vision
30
Via LLM Stats

Benchmark breakdown

Independent evaluation scores, normalized to 0–100.

GPQA Diamond
87
Humanity’s Last Exam
34
MMLU-Pro
SciCode
50
LiveCodeBench
MATH-500
AIME 2025
τ²-Bench (agentic)
94
Terminal-Bench Hard
43
IFBench
80
Via Artificial Analysis

Providers

Inference hosts serving this model — their own pricing and measured performance, cheapest input first.

ProviderIn $/1MOut $/1MThroughputLatencyStatus
Xiaomi$0.43$0.87active
DeepInfra$1.00$3.00active
Novita$2.00$6.00active
All providers

3 hosts · via LLM Stats

© 2026 NYSGPT2525 LLC