Mi:dm K 2.5 Pro Preview — benchmark results
Preview of KT's Mi:dm K 2.5 Pro, a 32B reasoning-focused Korean enterprise flagship strong in agentic tool use. Provider: KT. Released 2025-12-11. Access: API.
Unified ELO 1574 ± 29, rank #552 of 1776 rated models, from 45 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA AIME 2025 | 78.67 | Accuracy (%) | 74.1 |
| AA MMLU-Pro | 81.27 | Accuracy (%) | 73.7 |
| AA LiveCodeBench | 57.57 | Pass@1 (%) | 63.6 |
| AA Humanity's Last Exam | 8.8 | Accuracy (%) | 57.9 |
| AA GPQA Diamond | 72.22 | Accuracy (%) | 56.5 |
| AA Global-MMLU-Lite - Korean | 84.25 | Accuracy (%) | 54.8 |
| AA IFBench | 45.58 | Accuracy (%) | 52.5 |
| AA TAU-2 Bench | 49.42 | Accuracy (%) | 52.3 |
| AA Global-MMLU-Lite | 81.43 | Accuracy (%) | 47.9 |
| AA Omniscience - Software Engineering (SWE) - Swift | 36 | Accuracy (%) | 47.7 |
| AA Omniscience - Science, Engineering & Mathematics | 25.4 | Accuracy (%) | 46.8 |
| AA Omniscience - Software Engineering (SWE) - Java | 16 | Accuracy (%) | 46.7 |
Interactive version: theaggregate.ai/model?slug=mi-dm-k-2-5-pro-preview · How the rankings work · Data refreshed daily, snapshot 2026-07-22.