K2 Horizon 0.9B Pollard — Hardware Requirements & GPU Compatibility
Chat
K2 Horizon 0.9B Pollard is a 0.9B-parameter open language model from PollardWeights. At Q4_K_M it needs about 0.59 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 32K)Best on 24 GB (Q4_K_M · ~1111 tok/s · 32K)Macs from 8 GB (Q4_K_M · ~122 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run K2 Horizon 0.9B Pollard
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve PollardWeights/K2-Horizon-0.9B-Pollard --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 RTX 3060 12GB (10 GB headroom) · Q4_K_M
- Generation speed
- ~397 tok/s
- generation speed
- Cost per 1M output tokens
- $0.053
- per 1M output tokens
How Much VRAM Does K2 Horizon 0.9B Pollard Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 0.4 GB | 3.4 GB | 3-bit · i-quant small — Importance-weighted 3-bit, small | Fits: 32K context · ~1560 tok/s | |
| 0.4 GB | 3.4 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 32K context · ~1560 tok/s | |
| 0.5 GB | 3.5 GB | 3-bit medium — 3-bit medium quantization | Fits: 32K context · ~1365 tok/s | |
| 0.6 GB | 3.6 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 32K context · ~1111 tok/s | |
| 0.7 GB | 3.7 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 32K context · ~923 tok/s | |
| 0.8 GB | 3.8 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 32K context · ~799 tok/s | |
| 1.0 GB | 4 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 32K context · ~662 tok/s | |
| 2.0 GB | 5 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 32K context · ~331 tok/s |
Selected Q4_K_M
4-bit medium quantization2.0GBat 16K context
- Weights
- 0.1 GB
- KV cache
- 1.6 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 22.0 GB free · ~1111 tok/s
≥50 tok/s needs ≥45 GB/s · RTX 5090 ~1974, RTX 3090 Ti ~1111
Which Hardware Can Run K2 Horizon 0.9B Pollard?
Q4_K_M · 0.6 GBK2 Horizon 0.9B Pollard (Q4_K_M) requires 0.6 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 1+ GB is recommended. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 163 devices with unified memory can run K2 Horizon 0.9B Pollard, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB).
≤ 11 GB
≤ 11 GB tier →fits · 6.0 GB free · up to 32K ctx
16 GPUs · 1 Mac class · 2 device families
RTX 3080~838Simulate output on RTX 3080RTX 2080 Ti~679Simulate output on RTX 2080 TiRTX 3070 Ti~670Simulate output on RTX 3070 TiM3*8 GB~122Simulate output on iMac 24" M3 (8 GB)Orin Nano~112Simulate output on NVIDIA Jetson Orin Nano 8GB (Super)iPhone~81Simulate output on iPhone 1720 GB
fits · 18.0 GB free · up to 32K ctx
1 GPU
RX 7900 XT~814Simulate output on RX 7900 XT32–40 GB
fits · 30.0 GB free · up to 32K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~1974Simulate output on RTX 5090M5 Max36 GB~546Simulate output on Mac Studio M5 Max (36 GB)M532 GB~182Simulate output on MacBook Air 13" M5 (32 GB)Orin~226Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~149Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Similar models
1| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Qwen3.5 0.8BAlibaba | — | 0.9 GB | ~753 | same size |
Frequently Asked Questions
How much VRAM does K2 Horizon 0.9B Pollard need?
K2 Horizon 0.9B Pollard requires 0.6 GB of VRAM at Q4_K_M, or 2.0 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 0.9B × 4.8 bits ÷ 8 = 0.5 GB
KV Cache + Overhead ≈ 0.1 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.
VRAM usage by quantization
What's the best quantization for K2 Horizon 0.9B Pollard?
For K2 Horizon 0.9B Pollard, Q4_K_M (0.6 GB) offers the best balance of quality and VRAM usage. Q5_K_M (0.7 GB) provides better quality if you have the VRAM. The smallest option is IQ3_S at 0.4 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run K2 Horizon 0.9B Pollard on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 145 other Macs can run K2 Horizon 0.9B Pollard. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 0.6 GB of usable unified memory (RAM minus macOS overhead).
Can I run K2 Horizon 0.9B Pollard locally?
Yes — K2 Horizon 0.9B Pollard can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is K2 Horizon 0.9B Pollard?
At Q4_K_M, K2 Horizon 0.9B Pollard can reach ~8136 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~1111 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 ÷ 0.59 × 0.65 = ~8814 tok/s
Estimated speed at Q4_K_M (0.6 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 0.9B Pollard?
At Q4_K_M, the download is about 0.54 GB. The full-precision BF16 version is 1.80 GB. The smallest option (IQ3_S) is 0.38 GB.
Which GPUs can run K2 Horizon 0.9B Pollard?
52 consumer GPUs can run K2 Horizon 0.9B Pollard at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
Which devices can run K2 Horizon 0.9B Pollard?
163 devices with unified memory can run K2 Horizon 0.9B Pollard at Q4_K_M (0.6 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.