K2 Horizon 32B — Hardware Requirements & GPU Compatibility
Chat
K2 Horizon 32B is a 34.8B-parameter open language model from IFM. It supports a context window of up to 524,288 tokens. At BF16 it needs about 70.19 GB of VRAM — see which GPUs and Macs can run it below.
K2 Horizon 32B runs on a 96 GB+ Mac or a workstation GPU
Best local pick → Mac Studio (M3 Ultra, 96GB) · BF16 · 72.5 GB · ~8 tok/s
Macs from 96 GB (BF16 · ~8 tok/s)
Apache 2.0
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run K2 Horizon 32B
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
vllm serve IFM/K2-Horizon-32B --max-model-len 16384 --tensor-parallel-size 8vLLM repo with its own quantisation
Download: Weights on Hugging Face
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX PRO 6000 (96 GB) (23 GB headroom) · BF16
- Generation speed
- ~17 tok/s
- generation speed
- Cost per 1M output tokens
- $21.01
- per 1M output tokens
How Much VRAM Does K2 Horizon 32B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 70.2 GB | 75.2 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on H100 SXM 80 GB |
Selected BF16
Brain floating point 1672.5GBat 16K context
- Weights
- 69.6 GB
- KV cache
- 2.7 GB
- Overhead
- 0.3 GB
Server-class: ~72.5 GB at 16K context · one H100 SXM 80 GB
Or .
Which Hardware Can Run K2 Horizon 32B?
BF16 · 70.2 GBK2 Horizon 32B (BF16) requires 70.2 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 92+ GB is recommended. Using the full 524K context window can add up to 85.6 GB, bringing total usage to 155.8 GB. No consumer GPU has enough memory. 30 devices with unified memory can run K2 Horizon 32B, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 96GB).
96 GB
fits · 23.5 GB free · up to 128K ctx
3 GPUs · 2 Mac classes
RTX PRO 6000 Blackwell Max-Q Workstation Edition~17Simulate output on RTX PRO 6000 Blackwell Max-Q Workstation EditionRTX PRO 6000 Blackwell Workstation Edition~17Simulate output on RTX PRO 6000 Blackwell Workstation EditionRTX PRO 6000 Blackwell Server Edition~15Simulate output on RTX PRO 6000 Blackwell Server EditionM3 Ultra*96 GB~8Simulate output on Mac Studio (M3 Ultra, 96GB)M3 Max*96 GB~3Simulate output on MacBook Pro 14" M3 Max (96 GB)128–144 GB
fits · 55.5 GB free · up to 256K ctx
4 GPUs · 2 Mac classes · 4 device families
GH200 Grace Hopper Superchip~45Simulate output on GH200 Grace Hopper SuperchipH200 NVL~45Simulate output on H200 NVLH200 SXM~45Simulate output on H200 SXMM2 Ultra*128 GB~8Simulate output on Mac Pro M2 Ultra (128 GB)M5 Max128 GB~6Simulate output on Mac Studio M5 Max (128 GB)GB10~3Simulate output on ASUS Ascent GX10Jetson T5000 (Thor)~3Simulate output on NVIDIA Jetson AGX Thor Developer Kit
By VRAM tier · typical GPUs and Macs at BF16
Consumer GPUs (up to 32 GB) and Macs up to 64 GB: too small for this model (needs ≥ 72.5 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Qwen3.6 35B A3BAlibaba | 106.8 | 22 GB | ~151 | same size |
| MiniCPM V 4.7 35B A3BOpenBMB | — | 22.1 GB | ~169 | same size |
| Ornith 1.5 35B A3Bornith-ai | — | 22 GB | ~151 | same size |
| Agents A1InternScience | — | 21.5 GB | ~171 | same size |
| Intern S2 PreviewInternLM | — | 73.2 GB | —needs > 24 GB | same size |
| EXAONE 4.5 33BLGAI-EXAONE | — | 23.9 GB | ~33 | same size |
Specifications
- Publisher
- IFM
- Parameters
- 34.8B
- Architecture
- K2HorizonForCausalLM
- Context Length
- 524,288 tokens
- Vocabulary Size
- 250,624
- Release Date
- 2026-09-01
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does K2 Horizon 32B need?
K2 Horizon 32B requires 70.2 GB of VRAM at BF16. Full 524K context adds up to 85.6 GB (155.8 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 34.8B × 16 bits ÷ 8 = 69.6 GB
KV Cache + Overhead ≈ 0.6 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 86.2 GB (at full 524K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run K2 Horizon 32B?
No — K2 Horizon 32B requires at least 70.2 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
Can I run K2 Horizon 32B on a Mac?
Yes — Mac Studio (M3 Ultra, 96GB) and 20 other Macs can run K2 Horizon 32B. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 70.2 GB of usable unified memory (RAM minus macOS overhead).
Can I run K2 Horizon 32B locally?
Yes — K2 Horizon 32B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is K2 Horizon 32B?
At BF16, K2 Horizon 32B can reach ~68 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 70.2 × 0.65 = ~74 tok/s
Estimated speed at BF16 (70.2 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of K2 Horizon 32B?
At BF16, the download is about 69.56 GB.
Which GPUs can run K2 Horizon 32B?
No single consumer GPU has enough VRAM to run K2 Horizon 32B at BF16 (70.2 GB). Multi-GPU or professional hardware is required.
Which devices can run K2 Horizon 32B?
30 devices with unified memory can run K2 Horizon 32B at BF16 (70.2 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.