A.X K2 TPS calculator
Each card below is assessed against this model at the context length and minimum quality you choose. Speed is an estimate for a single request, calculated from the card's memory bandwidth and the size of the model once compressed.
Calculated for this model
818 cards we hold specifications for
Which GPUs can run A.X K2?
Set the inputs, read the answer
A longer conversation needs more memory, which can push this model off smaller cards.
Hides cards that would only fit the model by compressing it below this point.
0 cards match
Calculating| Needs | Quantisation | Fit | |||||
|---|---|---|---|---|---|---|---|
|
No card in our catalogue can run this model with these settings. |
|||||||
Speeds are estimates for a single request — one conversation at a time — calculated from memory bandwidth, model size and quantisation. Real throughput varies with the inference runtime and its version. Figures published by hardware vendors measure many simultaneous requests and are much higher.
On record
Full specification
Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.
Origin
Who built this model, where, and when it was published.
- Organisation
- SK Telecom
- Organisation type
- Industry
- Country
- Korea (Republic of)
- Published
- 29 July 2026
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Code generation, Language modeling/generation, Translation, Instruction interpretation, Mathematical reasoning, Chat, Language modeling, Language generation, Text autocompletion
Size
How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.
- Parameters
- 688B
- Training data
- tokens
MoE architecture, total parameters 688B, active parameters 33B (61 layers: 1 dense + 60 MoE; 256 routed experts + 1 shared, 8 active per token)
Approximately 8.5T tokens in total: ~8.2T in pre-training (Stage 1 general knowledge: 6.4T; Stage 2 high-quality reasoning: 1.4T; long-context stages 3A-3C: ~0.36T) and the remainder in post-training.
Training compute
The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.
- Training compute
- 1.8 × 10²⁴ FLOP
- How it was established
- Reported
~1.8e24 FLOP reported by SK Telecom via email (July 2026). Cross-check by operation counting: 6 * 33e9 active parameters * 8.5e12 tokens = 1.68e24 FLOP, consistent with the reported figure. Per the technical report, ~8.2T of the 8.5T tokens were used in pre-training (512 B200 GPUs, ~70 days), with the remainder in post-training.
The training run
What it physically took to train: which chips, how many, for how long, and what that drew from the wall.
- Wall-clock time
- 1,680 hours (70 days)
- Compute cost
- $12,400,000
Pre-training ran for approximately 70 days (~1,680 hours) on 512 NVIDIA B200 GPUs (64 nodes of 8), i.e. ~860,000 GPU-hours.
Availability
Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.
- Weights
- Open — downloadable
- Model access
- Open weights (unrestricted)
- Hugging Face
- skt
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Why it is tracked
- Training cost
- Record confidence
- Confident
Developer-reported training compute cost of ~$12.4M (2023 USD), communicated by SK Telecom via email (July 2026); exceeds the training cost threshold. Underlying cost calculation not yet provided.
Sources
Where this record came from and when it was last checked.
- Reference
- A.X K2 Technical Report
- Last updated
- 31 July 2026
What the numbers mean
The hardware side
A.X K2 reaches a parameter count of 688B. That is beyond what any single graphics card holds. Running it means either splitting it across several cards or renting hardware built for the job, and every card able to hold it alone is a datacentre part. The number that can: 0.
About this model
A.X K2 was published by SK Telecom, in the country recorded as Korea (Republic of), during July 2026. It comes out of an organisation categorised as industry.
It works in the domain of Language, and is recorded as performing the task of code generation, Language modeling/generation, Translation, Instruction interpretation, Mathematical reasoning, Chat, Language modeling, Language generation, Text autocompletion.
Published weights mean the model runs on your machine rather than someone else's, which is what makes the hardware question below answerable at all. On Hugging Face it is published under the organisation skt.
What went into building it
The training run consumed about 1.8 × 10²⁴ FLOP. That figure measures what producing the model cost, and has no bearing on how fast it answers.
Its inclusion criterion: training cost.
Step by step
How to choose a GPU for A.X K2
The table above has already assessed every card we hold specifications for against this model. Getting to your answer takes six steps.
-
01
Check what it needs before anything else
The table lists every card able to hold A.X K2. Capacity is the gate — a card either holds it or it does not.
-
02
Decide how long your conversations run
Longer conversations cost memory on top of what the weights need. Move the slider to your real working length before trusting any row for A.X K2.
-
03
Set a quality floor
The quantisation column varies by card, because a bigger card holds a more accurate copy. Setting a minimum quality drops the cards that only manage it by squeezing further than you would want, and holds the comparison at one level.
-
04
Rank by throughput rather than spec sheet
Ranking by tokens per second follows memory bandwidth rather than core counts, for A.X K2. It will not match a gaming ordering, because generation is bound by memory bandwidth.
-
05
Look at the headroom, not just the fit
A tight fit runs, but leaves nothing spare for a longer conversation, in the case of A.X K2. Comfortable means you can grow the context later. That difference matters more than a few tokens per second, so buy for comfortable if you expect to.
-
06
Check the card from the other side
Each card page repeats this sweep for every model we hold, answering what else the hardware is good for beyond A.X K2.
Answers
A.X K2 — common questions
A.X K2— where can I download it?
Its weights are published on Hugging Face, under the organisation skt. We do not host model files — this site calculates what hardware is needed to run them.
A.X K2— how much compute was used to train it?
Training consumed around 1.8 × 10²⁴ FLOP. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.
A.X K2— can I run it if it does not fit in my GPU?
It can be split between the card and system memory, but it generates painfully slowly that way. The nearest miss we calculate falls short by 110.8 GB. Every figure here assumes the whole model is resident on the card.
A.X K2— would two GPUs run it faster?
Two cards buy memory rather than speed, which matters only if one card cannot hold it. The number that can: 0. So a second card is rarely the answer here.
A.X K2— why does the quantisation differ between cards?
Because capacity varies, so does how hard it has to be squeezed. The number of distinct levels in the table above: 1. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.
A.X K2— how accurate are these speed estimates?
Every figure is derived from memory bandwidth and model size, not benchmarked, which is why each is published as a range rather than a single number. One example: the range beneath each figure. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.
A.X K2— is it open source?
Its weights are published, so it can be downloaded and run on your own hardware. Note that open weights is not the same as open source in the full sense — it says nothing about the training data, the training code, or the commercial terms attached.
A.X K2— how many parameters does it have?
It has a parameter count of 688B. MoE architecture, total parameters 688B, active parameters 33B (61 layers: 1 dense + 60 MoE; 256 routed experts + 1 shared, 8 active per token). That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.
A.X K2— who created it?
It was published by SK Telecom, based in Korea (Republic of), an organisation categorised as industry.
A.X K2— when was it released?
It was published in July 2026.
A.X K2— what is it used for?
It works in the domain of Language, and is recorded as handling the task of code generation, Language modeling/generation, Translation, Instruction interpretation, Mathematical reasoning, Chat, Language modeling, Language generation, Text autocompletion. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.
The other direction
Looking at it from the other side?
This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.