SKT A.X K2 Card - KMMLU-Pro: leaderboard
KMMLU-Pro professional-level Korean knowledge questions, run by SK Telecom. SKT A.X K2 model card benchmark table (July 2026), all models in thinking mode; accuracy in percent, higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Flash (Reasoning) | 78.8 | #155 (DeepSeek V4 Flash) |
| 2 | Qwen 3.5 397B A17B (Reasoning) | 78.4 | #154 (Qwen 3.5 397B A17B) |
| 3 | GLM-5.1 (Thinking) | 76.5 | #109 (GLM-5.1) |
| 4 | Kimi K2.6 (Thinking) | 73.4 | #113 (Kimi K2.6) |
| 5 | A.X K1 (Thinking) | 68.9 | #441 (A.X K1) |
| 6 | MiniMax-M2.7 | 66.6 | #270 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=skt-a-x-k2-card-kmmlu-pro · How It Works · Data refreshed daily, snapshot 2026-11-05.