Skip to main content

Xiaomi: MiMo-V2.5

Xiaomi: MiMo-V2.5 is xiaomi's model. On the un-leaked meo benchmark it ranks #11 of 22 with 50.4% accuracy and an Effective Value (𝕍) of 0.0003 β€” earning a correct answer for about $0.0026 in 94s each.

as of 2026-06-08

meo rank
#11
of 22
Accuracy
50.4%
overall
𝕍 index (top=100)
0.68
raw 𝕍 0.0003 Β· N=10
$ / correct
$0.0026
94s each

Per-domain accuracy

Theory of mind
100%
Regex automaton
100%
Tape machine
87%
CSP (unsatisfiable)
81%
Framework bias
67%
Perceptual illusions
52%
Long arithmetic
20%
Base-rate bias
7%
Logic / math / CS
5%
Watson-Glaser
0%
State tracking
0%

Efficiency

Cost per correct
$0.0026
Seconds per correct
94
Tokens per correct
8,813
Correct answers
107 / 251
Total benchmark cost
$0.28
Avg tokens / item
3,757

Model details

Lab
xiaomi
Released
2026-04-22
Weights
Open
Context
1.0M
Max output
131K
Knowledge cutoff
β€”
Price in
$0.14/M
Price out
$0.28/M
Modality
text+image+audio+video->text

Sources: openrouter Β· meo

FAQ

How accurate is Xiaomi: MiMo-V2.5?

Xiaomi: MiMo-V2.5 scores 50.4% overall on the meo benchmark (rank #11 of 22), an un-leaked private holdout graded objective-first.

What is Xiaomi: MiMo-V2.5's Effective Value?

Its Effective Value (𝕍) is 0.0003 at N=10, Ο‰=1, Ξ΄=1.5 β€” a fused metric of accuracy, speed, cost, and error-cascade. 𝕍 rankings shift with task depth.

How much does Xiaomi: MiMo-V2.5 cost per correct answer?

About $0.0026 per correct answer (8,813 tokens, 94 seconds each) at benchmark-time pricing via OpenRouter.

Is Xiaomi: MiMo-V2.5 open weights?

Yes β€” Xiaomi: MiMo-V2.5 is released with open weights.