LLMBOTTLENECK.COM
IFM / k2_horizon

K2-Horizon-375B-A23B

379.2 billion parameters, routing 8 of 192 experts per token.

Open in the calculator →

Architecturepublished data
Quick answer
K2-Horizon-375B-A23B needs about 230.5 GB at Q4_K_M — the format most people download — with 8,192 tokens of context. None of the 18 common devices listed below holds it entirely at this setting. No hardware for it? Rent a GPU that runs it, priced live.

Sized at 8,192 tokens of context with the whole model in device memory, across 18 common devices. Speeds are estimates, not benchmarks.

Official open weights (Apache-2.0), the flagship of IFM's K2-Horizon family, released 2026-09-01: a sparse mixture of experts that stores 375B parameters and runs 23B per token, native 524,288-token context, per the card. Licence declared on the repository: apache-2.0.

Weight-file size by format

FormatSizeRangeBasis
FP16754.7 GB747.1 GB – 762.2 GBreconstructed size
Q8_0401.0 GB397.0 GB – 405.1 GBreconstructed size
Q6_K309.7 GB306.6 GB – 312.8 GBreconstructed size
Q5_K_M267.4 GB264.8 GB – 270.1 GBreconstructed size
Q5_0259.8 GB257.2 GB – 262.4 GBreconstructed size
Q4_K_M227.7 GB225.4 GB – 230.0 GBreconstructed size
Q4_0212.8 GB210.7 GB – 215.0 GBreconstructed size
Q3_K_M186.5 GB172.5 GB – 200.5 GBreconstructed size
Q2_K137.3 GB131.1 GB – 143.5 GBreconstructed size

A file size is not the memory a run needs: the KV cache and the runtime reserve come on top, and the calculator adds both.

What the labels mean
published data
Read from a published source — a file's byte count, a model's configuration or a manufacturer's specification — or exact arithmetic on such values. Not a measurement on a machine.
reconstructed size
Weight size reconstructed from the pinned architecture, because no published file exists.
size range
Only a lower and an upper bound are claimed for this weight size.

Hardware ladder

Q4_K_M at 8,192 tokens · 0 of 18 devices hold it entirely
DeviceMemoryVerdictDecodeCalculator
NVIDIA GeForce RTX 5070230.5 GB of 12 GB59 layers on system RAM~4.2 tok/s with offloadOpen →
AMD Radeon™ RX 9070 XT230.5 GB of 16 GB58 layers on system RAM~4.3 tok/s with offloadOpen →
NVIDIA GeForce RTX 4080230.5 GB of 16 GB58 layers on system RAM~4.3 tok/s with offloadOpen →
NVIDIA GeForce RTX 5080230.5 GB of 16 GB58 layers on system RAM~4.3 tok/s with offloadOpen →
NVIDIA GeForce RTX 5070 Ti230.5 GB of 16 GB58 layers on system RAM~4.3 tok/s with offloadOpen →
NVIDIA GeForce RTX 5060 Ti 16GB230.5 GB of 16 GB58 layers on system RAM~4.2 tok/s with offloadOpen →
NVIDIA GeForce RTX 4090230.5 GB of 24 GB56 layers on system RAM~4.4 tok/s with offloadOpen →
AMD Radeon™ RX 7900 XTX230.5 GB of 24 GB56 layers on system RAM~4.4 tok/s with offloadOpen →
NVIDIA GeForce RTX 3090230.5 GB of 24 GB56 layers on system RAM~4.4 tok/s with offloadOpen →
NVIDIA GeForce RTX 5090230.5 GB of 32 GB54 layers on system RAM~4.6 tok/s with offloadOpen →
NVIDIA RTX 6000 Ada Generation230.5 GB of 48 GB49 layers on system RAM~5.0 tok/s with offloadOpen →
Apple M4 Pro230.5 GB of 64 GBdoes not fitnot calibratedOpen →
Apple M5 Pro230.5 GB of 64 GBdoes not fitnot calibratedOpen →
AMD Ryzen AI Max+ 395 with Radeon 8060S230.5 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M3 Max230.5 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M4 Max230.5 GB of 128 GBdoes not fitnot calibratedOpen →
NVIDIA DGX Spark230.5 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M2 Ultra230.5 GB of 192 GBdoes not fitnot calibratedOpen →

Decode is a calibrated estimate from the published calibration; capacity is the manufacturer's published ceiling, not guaranteed free memory. Fitting in memory is not the same as loading: whether the runtime and version you have supports this architecture and format on that machine has not been tested here. Offloaded rows assume there is enough system RAM for the overflow — the calculator checks that against the RAM you declare. Each “Open” link keeps this model, format and context.

Run it in the cloud

Rent a GPU that runs K2-Horizon-375B-A23B

Every rentable machine that holds the whole model, from Vast.ai and RunPod, with this site's speed estimate for each and the provider's own price, read live. Cheapest first; sort by cost per token or speed instead, or switch the format.

Prices · 18:30 UTC, 22 Sept
7 configurations in stock · 9 out of stock
MachineSpeedPer hourPer M tokensWhereRent
8× L40S384 GBrecommendedcheapestbest value~284 tok/sFaster than you read$3.73$3.64Vast.aiMarketplace · 99.9% reliableRent on Vast.ai
8× RTX PRO 5000384 GB~443 tok/sFaster than you read$6.81$4.27Vast.aiMarketplace · 98.3% reliableRent on Vast.ai
2× H200 NVL282 GB~395 tok/sFaster than you read$7.20$5.06Vast.aiMarketplace · 98.7% reliableRent on Vast.ai
1× B300270 GB~317 tok/sFaster than you read$7.89$6.91RunPodSecure CloudRent on RunPod
4× H100 SXM320 GBfastest~552 tok/sFaster than you read$9.07$4.56Vast.aiMarketplace · 99.9% reliableRunPod $13.96/h Rent on Vast.ai
2× H200282 GB~395 tok/sFaster than you read$9.18$6.44RunPodSecure CloudVast.ai $10.00/h Rent on RunPod
4× H100 PCIe320 GB~329 tok/sFaster than you read$9.59$8.07Vast.aiMarketplace · 98.3% reliableRent on Vast.ai
First time renting a GPU? How it works, in four steps
  1. Create an account and add credit. Both providers are prepaid: $10 is enough to try any card on this page for hours. A new RunPod account that arrives through a referral link gets a one-time $5 credit when it first adds $10.
  2. Pick the card and the count shown here, and a template: llama.cpp or vLLM for an OpenAI-compatible API, or one with a web chat if you only want to talk to the model. “How to launch it” below gives the exact image and command.
  3. Wait for the download. The weights are fetched on the machine; tens of gigabytes take a few minutes on a data-centre connection. You pay by the second from the moment it starts.
  4. Stop it when you are done. A running machine bills even when idle. A stopped one costs nothing per hour, though its disk is usually still billed until you delete it.
How to launch it

No Q4_K_M file of K2-Horizon-375B-A23B is catalogued here, so there is no exact command to vouch for. Start the machine from the llama.cpp server image ghcr.io/ggml-org/llama.cpp:server-cuda and point -hf at a Q4_K_M build from the publisher or a quantizer you trust — search Hugging Face for one. Check its file size against the memory figure on this page before you rent.

Every machine holds the whole model at Q4_K_M and 8,192 tokens of context, no offload. Speeds are this site’s single-stream decode estimates; prices are what each provider’s own API quoted, on-demand, for the whole machine. Vast hosts below 98% measured reliability are left out.

Referral links Vast.ai, RunPod and Novita pay us a share of what you spend if you sign up through these buttons. It costs you nothing, and it never decides an order or a recommendation: both are computed from the live price and the speed, and options that pay us nothing are listed and recommended on the same terms. How we rank

Buying a card instead, or paying by the token? Compare a year of K2-Horizon-375B-A23B three ways →

Under the hood

Architecture, read from the publisher’s file

The numbers every figure above is computed from, with the file they came from.

Architecture

✓ Architecture read from the published config.json

Retrieved 2026-09-24 at pinned commit 236f31838988; parameter count from the safetensors index at the same revision.

Architecture
k2_horizon
Layers
61
Hidden size
6,144
Attention heads
48
KV heads
8
Head dimension
128
Feed-forward width
16,384
Vocabulary
250,624
Context ceiling
524,288
RoPE theta
10,000,000
Experts
192
Experts / token
8
Expert width
1,792
Shared experts
1

Where the memory goes

tokenembedding× 61 decoder blocksGrouped-query attention48 query · 8 KV headsfull contextRouted experts8 of 192 per tokenall residentoutputprojectiongrows with contextcapacity ≠ traffic

Grouped-query attention shares each key/value head across 6 query heads, so the KV cache is 17% of what multi-head attention would need at the same context.

Open the exact official config.json ↗

More from IFM

K2-Horizon-0.9BK2-Horizon-3.7BK2-Horizon-7B

Use this answer in your own product

As JSON, from the API
curl -s https://llmbottleneck.com/v1/analyze \
  -H "Authorization: Bearer $LLMB_KEY" -H "content-type: application/json" \
  -d '{"model":"ifm-k2-horizon-375b-a23b","quantization":"Q4_K_M","context":8192,"hardware":"nvidia-geforce-rtx-4090"}'

Same engine, same evidence, every field sourced. Free key in one step →

On your page, as a widget
<script src="https://llmbottleneck.com/widget.js"
  data-model="ifm-k2-horizon-375b-a23b" data-quantization="Q4_K_M"
  data-hardware="nvidia-geforce-rtx-4090" data-context="8192"></script>

No key needed. Unbranded, with your own buy button, on Pro and Business →

Citing this page

LLM Bottleneck. “K2-Horizon-375B-A23B VRAM and hardware requirements.” Architecture from IFM/K2-Horizon-375B-A23B at revision 236f31838988, retrieved 2026-09-24. https://llmbottleneck.com/models/ifm-k2-horizon-375b-a23b

Every figure above is either the published value or a reconstruction whose measured error is on the accuracy page.

llmbottleneck
catalogue 2026-10-03models 327devices 135