Mi:dm K 2.5 Pro Preview

Korea Telecom·Released 2025-12-11

Limited comparison data25% coverage5/5 fields directly observedUpdated Jul 28, 2026, 6:00 PM

Benchmarks

MMLU-Pro

81.3%

GPQA Diamond

72.2%

HLE

8.8%

LiveCodeBench

57.6%

SciCode

29.7%

TerminalBench Hard

3.0%

MATH-500Not evaluated

AIMENot evaluated

AIME 2025

78.7%

IFBench

45.6%

Long Context Recall

11.0%

Tau2

49.4%

Market AverageTop Score

Claude Opus 5 (Adaptive Reasoning, Max Effort)

Anthropic

Claude Opus 5 (Adaptive Reasoning, Xhigh Effort)

Anthropic

Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)

Anthropic

GPT-5.6 Sol (max)

OpenAI

Claude Opus 5 (Adaptive Reasoning, High Effort)

Anthropic

GPT-5.6 Sol (xhigh)

OpenAI