Atlas

Models

← All models

Mi:dm K 2.5 Pro Preview

KT Corporation · 2025-12-11

Artificial Analysis identifies Mi:dm K 2.5 Pro Preview as a proprietary, text-only reasoning model but lists no API host or public weights. Atlas therefore represents the preview label as a benchmark entry rather than a released checkpoint.

Benchmark scores

BenchmarkScore
AA-LCR11.0
AA-Omniscience Index-62.0
AIME 202578.7
Artificial Analysis Omniscience Accuracy16.1
Artificial Analysis Omniscience Hallucination Rate93.1
CritPt0.0
GPQA Diamond72.2
Humanity's Last Exam (Text-Only)8.8
IFBench (Artificial Analysis)45.6
LiveCodeBench (Artificial Analysis)57.6
MMLU Pro81.3
SciCode (Artificial Analysis)29.7
Terminal-Bench Hard3.0
τ²-bench (Telecom)49.4
Loading Atlas data…