Skip to content
Model Pareto

A.X-K2

SK TelecomOpen weightsartificialanalysis.ai ↗
Input
—
Output
—
Speed
~83tok/sest.
Context
262Ktokens

Long-tail import from Artificial Analysis (2026-09-24); AA variant: A.X-K2. Price = AA median host price.

Rankings

Self-hostingEstimated

How we estimate →

What it would cost to run these open weights yourself on rented GPUs. No API sells this model, so this is its price on the chart.

Hardware
8× B200 180GB
FP8 weights
Throughput
~22,549 tok/s
many requests batched
Price
$0.25–2.21 /M tok
busy → light use
On your own machine
Mac Studio M5 Ultra 512GB (int4, ~51 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (688B params, 33B active per token) + 25% KV-cache headroom ≈ 860 GB
  • 8× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~22,549 output tok/s aggregate at batch 645 (bandwidth-bound); MoE compute scales with active params
  • Blended 3:1 input:output; prefill ~257,727 tok/s
  • Low = 75% utilization at the low GPU price; high = 20% at the high price

Benchmark scores

~ italic, dashed = no published score yet, estimated from related benchmarks.

Coding

~#14 of 21 (estimated)~71.7%
  1. #1Claude Opus 596.0%
  2. #2Claude Sonnet 585.2%
  3. #13Claude Haiku 4.573.3%
  4. ~#14A.X-K2Estimated71.7%
  5. #14Nemotron 3.5 Super VL71.1%
No published score — estimated from related benchmarks
~#23 of 24 (estimated)~50.9%
  1. #1Claude Fable 5.293.0%
  2. #2Claude Opus 5.589.9%
  3. #22Muse Glimmer51.2%
  4. ~#23A.X-K2Estimated50.9%
  5. #23Nex-N2.5-Mini43.8%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#38 of 46 (estimated)~45.3%
  1. #1Gemini 4 Argon78.8%
  2. #2DeepSeek V4.1 Flash 101777.0%
  3. #37Qwen3.8 Max (0902)51.0%
  4. ~#38A.X-K2Estimated45.3%
  5. #38Ripple-65B Preview44.3%
No published score — estimated from related benchmarks
LiveCodeBenchEstimated
~#15 of 28 (estimated)~64.3%
  1. #1Qwen3.8-Omni-Flash92.6%
  2. #2Qwen3.8 Flash-Next91.9%
  3. #14Maple-Preview75.1%
  4. ~#15A.X-K2Estimated64.3%
  5. #15Mistral Small 464.0%
No published score — estimated from related benchmarks
~#25 of 26 (estimated)~30.6
  1. #1Claude Opus 5.561.7
  2. #2Claude Opus 559.7
  3. #24GPT-6 Luna31.1
  4. ~#25A.X-K2Estimated30.6
  5. #25Muse Spark Open26.3
No published score — estimated from related benchmarks

Hard reasoning

Agents

~#12 of 15 (estimated)~64.6%
  1. #1Kimi K384.8%
  2. #2Qwen3.8 27B84.3%
  3. #11Muse Glimmer65.9%
  4. ~#12A.X-K2Estimated64.6%
  5. #12Nemotron 3.5 Super VL NVFP452.3%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~last of 5 (estimated)below measured range (<47.5%)
  1. #1GLM-5.3 Flash59.1%
  2. #2Kimi K358.3%
  3. #4GPT-5.547.5%
  4. ~#5A.X-K2Estimated<47.5%
No published score — estimated from related benchmarks
BrowseCompEstimated
~#19 of 22 (estimated)~63.3%
  1. #1Atria Dawn Preview92.5%
  2. #2GPT-6 Astra91.5%
  3. #18Nemotron 3.5 Bolt 74B-A9B70.0%
  4. ~#19A.X-K2Estimated63.3%
  5. #19Mistral Medium 3.548.6%
No published score — estimated from related benchmarks
~#31 of 92 (estimated)~1268 Elo
  1. #1Claude Opus 5.51866
  2. #2Muse Wonder Full1857
  3. #30GPT-5.51336
  4. ~#31A.X-K2Estimated1268
  5. #31MiniMax-M31242
No published score — estimated from related benchmarks
~#45 of 85 (estimated)~17.0%
  1. #1Claude Fable 5.276.4%
  2. #2Muse Wonder Full71.3%
  3. #44Qwen3.7 Plus17.4%
  4. ~#45A.X-K2Estimated17.0%
  5. #45Ling-3.0-flash-VL15.7%
No published score — estimated from related benchmarks