DA

danube3-4B

DanubeGGUF by h2oaiUpdated 2024-07-04
Download on Hugging Face
Parameters3.96B
Context8K
Layers24
Architecturellama
Smallest quant1.5 GB
Quick answer

danube3-4B needs about 3.6 GB of memory at Q4_K_M (a 2.4 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 8 quants, from 1.5 GB (Q2_K) to 7.9 GB (F16).

Danube3‑4B: 3.96 B parameter Llama‑style chat model

Danube3‑4B is a 3.96 B parameter Llama‑based chat model from H2O.ai. It supports a 8192‑token context window and is released under the Apache‑2.0 license. The model is available in multiple GGUF quantizations for llama.cpp.

Good at

  • Chatting with 8192‑token context
  • Low‑memory inference with Q2_K
  • Fast generation on GPUs

Limits

  • No vision support
  • Requires 8 k context for full performance
  • License is Apache‑2.0

Best for

  • Building lightweight chatbots
  • Running on edge devices
  • Rapid prototyping with llama.cpp

Who it’s for. Developers needing a small yet capable Llama‑style chat model. Suitable for projects that require a 8192‑token context window and can run on modest hardware.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K1.5 GB 2.8 GB
Fits Get file →
Q3_K_M1.9 GB 3.2 GB
Fits Get file →
Q4_K_M2.4 GB 3.6 GB
Fits Get file →
Q5_K_M2.8 GB 4.1 GB
Fits Get file →
Q6_K3.3 GB 4.5 GB
Fits Get file →
Q8_04.2 GB 5.5 GB
Fits Get file →
BF167.9 GB 9.2 GB
Too big Get file →
F167.9 GB 9.2 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (2.39 GB file, ~3.6 GB VRAM). If 4 GB VRAM is available, Q3_K_M (1.94 GB file, ~3.2 GB VRAM) is recommended. For the smallest footprint, Q2_K (1.51 GB file, ~2.8 GB VRAM) works on 2 GB VRAM setups.

What GPU do you need for danube3-4B?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is Q4_K_M (2.4 GB file, about 3.6 GB in use).
  • 6 GB VRAM: the best quant that fits is Q8_0 (4.2 GB file, about 5.5 GB in use).
  • 8 GB VRAM: the best quant that fits is Q8_0 (4.2 GB file, about 5.5 GB in use).
  • 12 GB VRAM: the best quant that fits is BF16 (7.9 GB file, about 9.2 GB in use).

Which danube3-4B quant should you choose?

Q4_K_M is the usual starting point for danube3-4B: it balances file size and quality well. Here it is a 2.4 GB file.

The Q8_0 file is 4.2 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is Q2_K at 1.5 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters3.96B
Layers24
Trained context8,192 tokens
KV cache (fp16)90 KB per token

Source

Quantized byh2oai
FamilyDanube
Licenseapache-2.0
Downloads (HF)3,001
Last updated2024-07-04

Frequently asked questions about danube3-4B

How much VRAM does danube3-4B need?

About 3.6 GB at Q4_K_M with an 8K context window. That is the 2.4 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run danube3-4B on an 8 GB GPU?

Yes, with Q8_0: the 4.2 GB file needs about 5.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of danube3-4B GGUF?

It depends on the quant: Q2_K is 1.5 GB, Q3_K_M is 1.9 GB, Q4_K_M is 2.4 GB, Q5_K_M is 2.8 GB, Q6_K is 3.3 GB, Q8_0 is 4.2 GB and 2 more.

What context length does danube3-4B support?

The GGUF header lists a trained context length of 8,192 tokens (8K). The KV cache grows with the context you use: about 0.8 GB at 8K and 0.8 GB at 8K.

Where can I download danube3-4B GGUF?

The files are on Hugging Face in the h2oai/h2o-danube3-4b-chat-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does danube3-4B use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Danube3‑4B on 8 GB VRAM?

Yes, the Q4_K_M quantization requires about 3.6 GB VRAM at 8 k context, well within 8 GB.

Does Danube3‑4B support vision?

No, the model has no vision capabilities.

Similar models

Browse all →

More from h2oai

All by h2oai →

More Danube models

All Danube →

Other small models

All small models →