A.X-K2
KRAFTONcc-by-nc-4.0audio

A.X-K2

Updated Aug 15, 2026
Parameters
21.2B
Context
8,192
License
cc-by-nc-4.0
Updated
Aug 15, 2026

Intelligence benchmarks

Artificial Analysis indexes - compared with the best open and proprietary models

Intelligence

35.0

AA Index

Agentic

25.7

AA Index

Intelligence Index - A.X-K2 vs. the field

Best open-weight models (you can run locally) and leading proprietary models for context.

Claude Opus 5 (max)
63.1
closed
Claude Fable 5 (with fallback)
62.1
closed
GPT-5.6 Sol (max)
60.9
closed
Grok 4.6 (high)
60.9
closed
Kimi K3
59.7
open
Qwen3.8 Max
58.1
closed
Qwen3.8 2.4T A95B
57.7
open
A.X-K2
35
open

Agentic Index comparison

Claude Opus 5 (max)
59.2
closed
Grok 4.6 (high)
58.7
closed
Qwen3.8 Max
58.4
closed
GPT-5.6 Sol (max)
57.8
closed
Qwen3.8 2.4T A95B
57.1
open
Claude Fable 5 (with fallback)
56.6
closed
Kimi K3
54.3
open
A.X-K2
25.7
open

Benchmark data from Artificial Analysis · updated 2026-08-15.

Standard benchmarks

Performance across standard evaluations

BenchmarkScore
GPQA85.7

Will it run on your hardware?

Pick your GPU memory - see which quantizations fit, and the cheapest card for the rest

Runs on your 24 GB - best at Q5_K_M
2 of 4 quantizations fit A.X-K2 with real runtime overhead.
Q4_K_M
12 GB
Q5_K_M
15 GB
fits tight too big

Need an exact figure for your context length? Use the VRAM calculator.

Run it locally

Copy-paste - running in under a minute

vLLMOpenAI-compatible API
vllm serve KRAFTON/A.X-K2-Raon-Speech-21B-A3B

New to this? Start with Ollama · serve to many users with vLLM.

Deep dive

Notes, sources, and the full write-up

A.X-K2 is a 21.2B-parameter cc-by-nc-4.0 model from SK Telecom. It scores 35 on the Artificial Analysis Intelligence Index. At Q4_K_M it needs roughly 12 GB of VRAM, placing it in the 8–12 GB GPU hardware tier.

Benchmarks

Artificial Analysis Intelligence Index - A.X-K2 vs. leading closed models:

ModelIntelligenceCodingGPQA
A.X-K235-85.7
Claude Opus 5 (max)63.1-93.2
Claude Fable 5 (with fallback)62.1-92.6
GPT-5.6 Sol (max)60.9-94.1
Grok 4.6 (high)60.9-94.9
Qwen3.8 Max58.1-92.7

Source: Artificial Analysis (2026-08-13).

Popularity

A.X-K2 has 5,454 downloads in the last month on HuggingFace and 96 likes.

Frequently asked

Quick answers to common questions

How much VRAM does A.X-K2 need?

A.X-K2 with 21.2B parameters needs approximately 12 GB at Q4_K_M quantization. Use our VRAM calculator for an exact estimate.

Is A.X-K2 better than other KRAFTON models?

A.X-K2 has 21.2B parameters with 8,192 context - a strong choice for general use.

What license is A.X-K2 under?

A.X-K2 is released under the cc-by-nc-4.0 license, making it suitable for most commercial and personal projects.

What hardware runs A.X-K2 well?

With 21.2B parameters, A.X-K2 requires adequate VRAM. High-end GPUs like the RTX 4090 (24GB), RTX 5090 (32GB), or Mac Studio with unified memory are good options. Check our hardware directory for specific recommendations.

What is the best quantization for A.X-K2?

Q4_K_M is the recommended sweet spot - ~98% of FP16 quality at ~27% of the size. Q5_K_M (~15 GB) is an option if you have spare VRAM. Use our VRAM calculator to compare.

What models compete with A.X-K2?

A.X-K2 competes with other models in its class. Browse our model directory for comparisons, benchmarks, and community reviews to find the best fit.

Nearby options

Similar models and compatible hardware by spec

Comments coming soon

Configure NEXT_PUBLIC_GISCUS_REPO_ID and NEXT_PUBLIC_GISCUS_CATEGORY_ID at giscus.app to enable.