Kappa 20B 131k — Hardware Requirements & GPU Compatibility
Chat
400 downloads
12 likesSpecifications
- Publisher
- eousphoros
- Parameters
- 20.9B
- Architecture
- GptOssForCausalLM
- Context Length
- 131,072 tokens
- Vocabulary Size
- 201,088
- Release Date
- 2026-02-28
- License
- Other
Get Started
HuggingFace
How Much VRAM Does Kappa 20B 131k Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16 | 16.00 | 42.2 GB | 46.7 GB | 41.83 GB | Brain floating point 16 — preferred for training |
Which GPUs Can Run Kappa 20B 131k?
BF16 · 42.2 GBKappa 20B 131k (BF16) requires 42.2 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 55+ GB is recommended. Using the full 131K context window can add up to 4.5 GB, bringing total usage to 46.7 GB. No single GPU has enough memory — multi-GPU or cluster setups are needed.
Which Devices Can Run Kappa 20B 131k?
BF16 · 42.2 GB11 devices with unified memory can run Kappa 20B 131k, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, MacBook Pro 16" M4 Max (48 GB).
Runs great
— Plenty of headroom
Decent
— Enough memory, may be tightRelated Models
Frequently Asked Questions
- How much VRAM does Kappa 20B 131k need?
Kappa 20B 131k requires 42.2 GB of VRAM at BF16. Full 131K context adds up to 4.5 GB (46.7 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 20.9B × 16 bits ÷ 8 = 41.8 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 4.9 GB (at full 131K context)
VRAM usage by quantization
BF1642.2 GBBF16 + full context46.7 GB- Can NVIDIA GeForce RTX 5090 run Kappa 20B 131k?
No — Kappa 20B 131k requires at least 42.2 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
- Can I run Kappa 20B 131k on a Mac?
Kappa 20B 131k requires at least 42.2 GB at BF16, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run Kappa 20B 131k locally?
Yes — Kappa 20B 131k can run locally on consumer hardware. At BF16 quantization it needs 42.2 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Kappa 20B 131k?
At BF16, Kappa 20B 131k can reach ~69 tok/s on AMD Instinct MI300X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 42.2 × 0.55 = ~69 tok/s
Estimated speed at BF16 (42.2 GB)
AMD Instinct MI300X~69 tok/sNVIDIA H100 SXM~52 tok/sAMD Instinct MI250X~43 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Kappa 20B 131k?
At BF16, the download is about 41.83 GB.
- Which GPUs can run Kappa 20B 131k?
No single consumer GPU has enough VRAM to run Kappa 20B 131k at BF16 (42.2 GB). Multi-GPU or professional hardware is required.
- Which devices can run Kappa 20B 131k?
11 devices with unified memory can run Kappa 20B 131k at BF16 (42.2 GB), including Mac Mini M4 Pro (48 GB), Mac Pro M2 Ultra (192 GB), Mac Studio M2 Ultra (192 GB), Mac Studio M4 Max (128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.