Skip to content

K2 Horizon 375B A23B VRAM requirements

How much VRAM K2 Horizon 375B A23B needs
4-bit weights: 214 GBWith 8K context: 217 GBWith 32K context: 223 GB8-bit, 8K context: 378 GB
Parameters
379B
Active per token
23B
Longest context
512K
Released
Sep 2026
License
Apache 2.0
Strongest at
Not rated

How much VRAM K2 Horizon 375B A23B needs

Memory K2 Horizon 375B A23B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit706 GB709 GB715 GB738 GB
8-bit375 GB378 GB384 GB407 GB
6-bit290 GB293 GB298 GB321 GB
5-bit251 GB254 GB260 GB283 GB
4-bit214 GB217 GB223 GB246 GB
3-bit173 GB176 GB181 GB204 GB
2-bit132 GB135 GB141 GB164 GB
1.6-bit72.8 GB75.9 GB81.6 GB104 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run K2 Horizon 375B A23B

The best way K2 Horizon 375B A23B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: K2 Horizon 375B A23B on RTX 4060 report
RTX 5060Too bigPDF: K2 Horizon 375B A23B on RTX 5060 report
RTX 3070Too bigPDF: K2 Horizon 375B A23B on RTX 3070 report
RTX 3080 10 GBToo bigPDF: K2 Horizon 375B A23B on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: K2 Horizon 375B A23B on RTX 3060 12 GB report
Arc B580Too bigPDF: K2 Horizon 375B A23B on Arc B580 report
RTX 4070Too bigPDF: K2 Horizon 375B A23B on RTX 4070 report
RTX 4070 SuperToo bigPDF: K2 Horizon 375B A23B on RTX 4070 Super report
RTX 5070Too bigPDF: K2 Horizon 375B A23B on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: K2 Horizon 375B A23B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: K2 Horizon 375B A23B on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: K2 Horizon 375B A23B on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: K2 Horizon 375B A23B on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: K2 Horizon 375B A23B on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: K2 Horizon 375B A23B on RTX 4080 Super report
RTX 5070 TiToo bigPDF: K2 Horizon 375B A23B on RTX 5070 Ti report
RTX 5080Too bigPDF: K2 Horizon 375B A23B on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: K2 Horizon 375B A23B on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: K2 Horizon 375B A23B on RTX 3090 report
RTX 4090Too bigPDF: K2 Horizon 375B A23B on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: K2 Horizon 375B A23B on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: K2 Horizon 375B A23B on RTX 5090 report
2× RTX 3090Too bigPDF: K2 Horizon 375B A23B on 2× RTX 3090 report
2× RTX 4090Too bigPDF: K2 Horizon 375B A23B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: K2 Horizon 375B A23B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: K2 Horizon 375B A23B on RTX 6000 Ada report
L40SToo bigPDF: K2 Horizon 375B A23B on L40S report
2× RTX 5090Too bigPDF: K2 Horizon 375B A23B on 2× RTX 5090 report
4× RTX 3090Too bigPDF: K2 Horizon 375B A23B on 4× RTX 3090 report
RTX PRO 6000 BlackwellToo bigPDF: K2 Horizon 375B A23B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: K2 Horizon 375B A23B on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: K2 Horizon 375B A23B on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: K2 Horizon 375B A23B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: K2 Horizon 375B A23B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: K2 Horizon 375B A23B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: K2 Horizon 375B A23B on Mac mini M4 Pro 64 GB report
M4 Max 128 GBToo bigPDF: K2 Horizon 375B A23B on M4 Max 128 GB report
M5 Max 128 GBToo bigPDF: K2 Horizon 375B A23B on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: K2 Horizon 375B A23B on M3 Ultra 96 GB report
M3 Ultra 256 GB3-bit36K34 tok/s13 sPDF: K2 Horizon 375B A23B on M3 Ultra 256 GB report
M3 Ultra 512 GB6-bit288K23 tok/s13 sPDF: K2 Horizon 375B A23B on M3 Ultra 512 GB report
M5 Ultra 512 GB6-bit288K31 tok/s4.1 sPDF: K2 Horizon 375B A23B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: K2 Horizon 375B A23B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GBToo bigPDF: K2 Horizon 375B A23B on Ryzen AI Max+ 395 128 GB report
RTX SparkToo bigPDF: K2 Horizon 375B A23B on RTX Spark report
DGX SparkToo bigPDF: K2 Horizon 375B A23B on DGX Spark report
Dual DGX Spark4-bit52K13 tok/s2.7 sPDF: K2 Horizon 375B A23B on Dual DGX Spark report
DGX Station GB3004-bit96K155 tok/s0.4 sPDF: K2 Horizon 375B A23B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell3-bit32K79 tok/s0.6 sPDF: K2 Horizon 375B A23B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell6-bit288K53 tok/s0.3 sPDF: K2 Horizon 375B A23B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit512K43 tok/s0.2 sPDF: K2 Horizon 375B A23B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: K2 Horizon 375B A23B on A100 80 GB SXM report
H100 SXMToo bigPDF: K2 Horizon 375B A23B on H100 SXM report
4× H200 NVL8-bit512K91 tok/s0.3 sPDF: K2 Horizon 375B A23B on 4× H200 NVL report
8× H200 NVL8-bit512K91 tok/s0.2 sPDF: K2 Horizon 375B A23B on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf bartowski/K2-Horizon-375B-A23B-GGUF
vLLM
vllm serve IFM/K2-Horizon-375B-A23B

Questions

How much VRAM does K2 Horizon 375B A23B need?

K2 Horizon 375B A23B needs 214 GB for its weights at 4-bit, 217 GB with an 8K context and 223 GB with 32K. At 8-bit, 378 GB.

What is the minimum VRAM for K2 Horizon 375B A23B?

About 135 GB at 2-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the M3 Ultra 256 GB.

How big is K2 Horizon 375B A23B?

379B parameters (23B active per token): 706 GB of weights at 16-bit, 214 GB at 4-bit.

Can K2 Horizon 375B A23B run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the M3 Ultra 256 GB.

More models