K2 Horizon 3.7B — Hardware Requirements & GPU Compatibility
Chat
K2 Horizon 3.7B is a 5.1B-parameter open language model from IFM. It supports a context window of up to 524,288 tokens. At BF16 it needs about 10.61 GB of VRAM — see which GPUs and Macs can run it below.
Runs well on a 24 GB GPU
Runs from 12 GB (BF16 · 16K)Best on 24 GB (BF16 · ~62 tok/s · 128K)Macs from 16 GB (BF16 · ~13 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run K2 Horizon 3.7B
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve IFM/K2-Horizon-3.7B --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on Tesla P40 (24 GB) (12 GB headroom) · BF16
- Generation speed
- ~21 tok/s
- generation speed
- Cost per 1M output tokens
- $1.59
- per 1M output tokens
How Much VRAM Does K2 Horizon 3.7B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 10.6 GB | 13.4 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 128K context · ~62 tok/s |
Selected BF16
Brain floating point 1611.9GBat 16K context
- Weights
- 10.1 GB
- KV cache
- 1.5 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 12.1 GB free · ~62 tok/s
≥50 tok/s needs ≥815 GB/s · RTX 5090 ~110, RTX 3090 Ti ~62
Which Hardware Can Run K2 Horizon 3.7B?
BF16 · 10.6 GBK2 Horizon 3.7B (BF16) requires 10.6 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 14+ GB is recommended. Using the full 524K context window can add up to 48.1 GB, bringing total usage to 58.7 GB. 38 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3080 Ti. 150 devices with unified memory can run K2 Horizon 3.7B, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Mini M2 Pro (16 GB).
12 GB
12 GB tier →fits · 0.1 GB free · up to 16K ctx
10 GPUs
RTX 3080 Ti~56Simulate output on RTX 3080 TiRTX 5070~41Simulate output on RTX 5070RTX 4070~31Simulate output on RTX 407020 GB
fits · 8.1 GB free · up to 64K ctx
1 GPU
RX 7900 XT~45Simulate output on RX 7900 XT32–40 GB
fits · 20.1 GB free · up to 128K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~110Simulate output on RTX 5090M5 Max36 GB~30Simulate output on Mac Studio M5 Max (36 GB)M532 GB~10Simulate output on MacBook Air 13" M5 (32 GB)Orin~13Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~8Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at BF16
Similar models
3| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Gemma 4 E2BGoogle | — | 3.5 GB | ~191 | same size |
| Molmo2 4BAllen AI | — | 4.7 GB | ~193 | same size |
| Gemma 3n E2B ITGoogle | — | 5.9 GB | ~183 | same size |
Specifications
- Publisher
- IFM
- Parameters
- 5.1B
- 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 3.7B need?
K2 Horizon 3.7B requires 10.6 GB of VRAM at BF16. Full 524K context adds up to 48.1 GB (58.7 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 5.1B × 16 bits ÷ 8 = 10.1 GB
KV Cache + Overhead ≈ 0.5 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 ≈ 48.6 GB (at full 524K context)
VRAM usage by quantization
Can I run K2 Horizon 3.7B on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run K2 Horizon 3.7B. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 10.6 GB of usable unified memory (RAM minus macOS overhead).
Can I run K2 Horizon 3.7B locally?
Yes — K2 Horizon 3.7B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is K2 Horizon 3.7B?
At BF16, K2 Horizon 3.7B can reach ~452 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~62 tok/s. 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 ÷ 10.61 × 0.65 = ~490 tok/s
Estimated speed at BF16 (10.6 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 3.7B?
At BF16, the download is about 10.12 GB.
Which GPUs can run K2 Horizon 3.7B?
38 consumer GPUs can run K2 Horizon 3.7B at BF16. Top options include AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 6900 XT, AMD Radeon RX 6700 XT. 26 GPUs have plenty of headroom for comfortable inference.
Which devices can run K2 Horizon 3.7B?
150 devices with unified memory can run K2 Horizon 3.7B at BF16 (10.6 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (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.