What are the hardware requirements for H2o Danube 1.8B Sft?

H2o Danube 1.8B Sft requires approximately 3.7 GB of VRAM and supports a context window of 16K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed H2o Danube 1.8B Sft and how large is it?

H2o Danube 1.8B Sft is developed by h2oai, a model with 1.8b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does H2o Danube 1.8B Sft perform on standard benchmarks?

H2o Danube 1.8B Sft has the following published scores: MMLU 33.75. Compare against reference models on this page or on the LLM Explorer leaderboards.

H2o Danube 1.8B Sft by h2oai — VRAM 3.7GB, 16K context

Name: H2o Danube 1.8B Sft
Author: h2oai

H2o Danube 1.8B Sft is an open-source language model by h2oai. Features: 1.8b LLM, VRAM: 3.7GB, Context: 16K, License: apache-2.0, LLM Explorer Score: 0.14, Arc: 40.2, HellaSwag: 67.3, MMLU: 33.8, GSM8K: 15.1.

Arxiv:2401.16818 Conversational Dataset:huggingfaceh4/ultracha... Dataset:meta-math/metamathqa Dataset:open-orca/openorca Dataset:openassistant/oasst2 En Endpoints compatible Gpt H2o-llmstudio Mistral Region:us Safetensors

Model Card on HF 🤗: https://huggingface.co/h2oai/h2o-danube-1.8b-sft

H2o Danube 1.8B Sft Parameters and Internals

Model Type

chat, fine-tuned

Use Cases

Areas:

Research, Commercial applications

Limitations:

Incorrect, nonsensical or irrelevant responses

Considerations:

Users should ensure ethical use and be aware of potential biases in the generated outputs.

Additional Notes

The model architecture details and usage guidelines are provided in the model card and technical report.

Training Details

Data Sources:

Open-Orca/OpenOrca, OpenAssistant/oasst2, HuggingFaceH4/ultrachat_200k, meta-math/MetaMathQA

Methodology:

The model was trained using H2O LLM Studio with sliding window attention from mistral.

Context Length:

16384

Model Architecture:

MistralForCausalLM incorporating Llama 2 architecture with sliding window attention.

Input Output

Performance Tips:

Quantization and sharding are supported in the transformers library for efficient usage.

LLM Name	H2o Danube 1.8B Sft
Repository 🤗	https://huggingface.co/h2oai/h2o-danube-1.8b-sft
Model Size	1.8b
Required VRAM	3.7 GB
Updated	2026-08-05
Maintainer	h2oai
Model Type	mistral
Model Files	3.7 GB
Supported Languages	en
Model Architecture	MistralForCausalLM
License	apache-2.0
Context Length	16384
Model Max Length	16384
Transformers Version	4.36.1
Tokenizer Class	LlamaTokenizer
Padding Token	<unk>
Vocabulary Size	32000
Torch Data Type	bfloat16

Rank the H2o Danube 1.8B Sft capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to H2o Danube 1.8B Sft

Best Alternatives	Context / RAM	Downloads	Likes
H2o Danube 1.8B Chat	16K / 3.7 GB	267	55
H2o Danube 1.8B Base	16K / 3.7 GB	132	43
Cypher Mini 1.8B	16K / 3.7 GB	83	2
PixieZehirNano	16K / 3.7 GB	10	0
Cypher CoT 1.8B	16K / 3.7 GB	5	1
...1.8B Chat Sft Merge Fourier V1	16K / 7.3 GB	3	1
H2o Danube2 1.8B Base	8K / 3.7 GB	346	47
H2o Danube2 1.8B Chat	8K / 3.7 GB	222	62
H2o Danube2 1.8B Sft	8K / 3.7 GB	109	6
Binary Clumsy Bear	8K / 7.3 GB	5	0

Note: green Score (e.g. "73.2") means that the model is better than h2oai/h2o-danube-1.8b-sft.