whatmodel.sh / A.X K2 (688B-A33B)
A.X K2 (688B-A33B)
↗ share
+ contribute
A.X K2 (688B-A33B) is an open-weights language model from SK Telecom, released July 2026. Its published size is 691.69B parameters (33B active per token) and a 262k-token context window. Licence: apache-2.0.
Catalogue identity notes
skt/A.X-K2 @2287ca4 (created 2026-07-28, licence apache-2.0). SK Telecom A.X K2 (July 2026), 688B MoE with 33B active, Korean/English; thinking and non-thinking modes. Rows from SKT's CC BY technical report (own runs of A.X K2/K1 and five open-weight baselines).
Model information sources
Release date ↗ · Weights licence ↗
HF safetensors metadata 691,690,231,040 parameters; 33B active per the maker card.
$ how it compares
The LLMs ranked either side of A.X K2 (688B-A33B), out of 828 with a score, then more from SK Telecom.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 99 |  Qwen3.6 27B | 69±1 | api $1.20/M +local | ||
| 100 |  Ring-2.6-1T | 69±2 | comparison only | ||
| 101 | A.X K2 (688B-A33B) | 69±1 | local | ||
| 102 |  o3 | 69±1 | api $3.50/M | ||
| 103 |  MiMo-V2.5 (310B-A15B) | 69±1 | api $0.21/M +local | ||
| more from SK Telecom | |||||
| 208 | A.X K1 (519B-A33B) | 62±1 | local | ||
$ what it's good at
coding▁▁▁▃▆██▇▆▄▃▁ #76 of 458 * · 18% of its score · 2 tests
Beats GPT-5.2, Grok 4.20 (0309 reasoning), GPT-5.4 mini here, though they rank higher overall. Trails GLM-4.7, Gemini 3.5 Flash-Lite, MiMo-V2.5 (310B-A15B) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LiveCodeBench v6 (Feb–May 2025) lab | 84.0% ✎ | ▄··█····▄▄·█ beats 4/6 | 8.9% |
| SciCode (level unspecified) lab | 41.0% ✎ | ▄·····▄█▄▄·▄ beats 2/6 | 8.6% |
math▁▂▃▅▆▅▄█▄▁▁▁ #43 of 398 * · 7.7% of its score · 1 test
Beats Grok 4.7, GLM-5.3-Flash, Gemini 3.6 Flash here, though they rank higher overall.
other tests4 not in a task · 74% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond lab | 85.6% ✎ | ▄·····▄█▄▄·▄ beats 1/6 | 33% |
| Global-MMLU-Lite lab | 87.6% ✎ | ▃······▅··█▃ beats 2/6 | 17% |
| Humanity's Last Exam lab | 27.8% ✎ | ▂········█·▂ beats 3/6 | 15% |
| IFBench (aggregation unspecified) lab | 75.9% ✎ | ▃······█▃▃·▃ beats 2/6 | 9.5% |
$ where to run it
Other ways to run it
- Run it as published · needs ~1383.4 GB, NVIDIA only · from huggingface.co/skt
vllm serve skt/A.X-K2 --tensor-parallel-size 8
Weights and documentation: huggingface.co/skt ↗
$ how the score was worked out · A.X K2 Technical Report (SK T… 99%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 A.X K2 Technical Report (SK Telecom), arXiv 2608.30181v1 · lab-reported | 7 | 99% |
| prior · pull toward the middle | 1% | |
| weighted average | = 69.3 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json