Compare Models

MiMo V2.5 Pro

MiMo V2.5 Pro is an open-weight model from Xiaomi, released April 22, 2026. It ranks #40 of 45 models on the Vals Index with 33.84%. Its best result is #40 of 107 on MedScribe.

Release Date: Apr 22, 2026

Developer Xiaomi  🇨🇳
Context Window 1M
Max Output Tokens 128k
Token Costs (in/out) $0.43/0.87
Weights Open
Input Modalities

Accuracy

33.84 % ± 1.17

Cost / Test (Vals Index)

$ 0.162

Latency

34 min 33 s

Vals Index
BenchmarksAccuracyRankings
Vals Index

33.84%

±1.17
40/45
Code Migration

21.56%

±4.13
47/74
EMB

55.23%

±3.11
46/72
Finance Agent (v2)

41.50%

±1.19
57/76
Legal Research Bench

15.87%

±2.54
55/75
MedCode

32.48%

±1.91
87/105
MedScribe

83.73%

±2.06
40/107
ProofBench v1.1

22.00%

±4.16
37/48
Tax Agent Bench

34.79%

±2.62
56/67
TaxEval v2

73.79%

±0.86
46/145
Vibe Code Bench v1.1

34.11%

±4.53
66/110
Harvey's Legal Agent Benchmark

2.08%

±0.83
44/76
GPQA Diamond

82.58%

±2.47
61/138
LiveCodeBench

81.35%

±1.07
64/143
LegalBench

77.13%

±0.55
111/149
MMLU Pro

84.59%

±0.45
62/138
SWE-bench

74.00%

±1.96
50/88
Terminal-Bench 4.0

0.51%

±0.51
43/45
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Xiaomi
Temperature: 1
Top P: 0.95
Top K: Default
Max Output Tokens: 128,000

Compared with

MiMo V2.5 Pro vs Gemini 3.1 Pro Preview (02/26) 18 shared benchmarks MiMo V2.5 Pro vs GPT 5.4 Mini 14 shared benchmarks MiMo V2.5 Pro vs Kimi K2.6 15 shared benchmarks MiMo V2.5 Pro vs MiniMax-M3 18 shared benchmarks MiMo V2.5 Pro vs Qwen 3.7 Plus 7 shared benchmarks MiMo V2.5 Pro vs DeepSeek V4 18 shared benchmarks MiMo V2.5 Pro vs Qwen 3.7 Max 15 shared benchmarks MiMo V2.5 Pro vs Gemini 3.5 Flash Lite 15 shared benchmarks MiMo V2.5 Pro vs GLM 5.3 Flash 15 shared benchmarks MiMo V2.5 Pro vs GPT 5.4 Nano 15 shared benchmarks