A.X K2 measured on 7 of 42 Agents & Tool Use benchmarks7
BenchmarkByScoreRank
Agents6
Documents1
No benchmark matches
—per task?
self-host onlyprice
—output speed?
262Kcontext
Alternatives?
ModelScoreCostPer taskWhat changes
ModelScoreCostPer taskWhat changes
Nothing here yet.
ModelScoreCostPer taskWhat changes
ModelScoreCostPer taskWhat changes
Run it yourself
estimateFits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✗ 48 GB GPU✗ 80 GB GPU
LicenceApache-2.0free for commercial use
Benchmarks
in Agents & Tool UseTerminal-Bench 2.139#87
GDPval-AA (work tasks) 2.1959.9#76
τ³-Banking · AA16.1#65
AutomationBench-AA2.5#96
AA-Briefcase 1.1 · Elo737.8#83
FAQ
How much does A.X K2 cost?›
There is no per-unit API price: self-host only.
Is A.X K2 free to use?›
There is no free API. The weights are open, so you can run it yourself.
How good is A.X K2, and what is it best at?›
#85 of 336 in Chat & Knowledge, #77 of 134 in Coding, #96 of 135 in Agents & Tool Use, #78 of 135 in Vision & Documents. Strongest at: Science.
Is A.X K2 better than DeepSeek V3.1 Terminus?›
On AutomationBench-AA, A.X K2 scores 2.5 vs 2.7 — about the same.
What is the context window of A.X K2?›
262K tokens.
When was A.X K2 released? Is there a newer version?›
Released 2026-08-12. It is the current version in its line.
Is A.X K2 open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run A.X K2 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can A.X K2 read images?›
No, it accepts text only.
What is A.X K2?›
A.X K2 is an open-weights language model from SK Telecom, released Aug 2026.
What are the best alternatives to A.X K2?›
GPT-6 Luna (better), DeepSeek V4.1 Flash 1017 (open weights).