What are the hardware requirements for Danube2 Upscale 1.7?

Danube2 Upscale 1.7 requires approximately 4.5 GB of VRAM and supports a context window of 8K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Danube2 Upscale 1.7 and how large is it?

Danube2 Upscale 1.7 is developed by Lambent, a model with 2.2b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Danube2 Upscale 1.7 perform on standard benchmarks?

Danube2 Upscale 1.7 has the following published scores: MMLU 40.38. Compare against reference models on this page or on the LLM Explorer leaderboards.

Danube2 Upscale 1.7 by Lambent — VRAM 4.5GB, 8K context

Name: Danube2 Upscale 1.7
Author: Lambent

Danube2 Upscale 1.7 is an open-source language model by Lambent. Features: 2.2b LLM, VRAM: 4.5GB, Context: 8K, License: apache-2.0, Merged, LLM Explorer Score: 0.15, Arc: 43.3, HellaSwag: 69, MMLU: 40.4, GSM8K: 11.3.

Merged Model Arxiv:2203.05482 Base model:lambent/danube2-ups... Base model:lambent/danube2-ups... Base model:lambent/danube2-ups... Base model:lambent/danube2-ups... Dataset:huggingfacetb/cosmoped... Dataset:nampdn-ai/tiny-bridged... Dataset:severian/internal-know... Dataset:severian/internal-know... Dataset:sordonia/redpajama-sam... Dataset:teknium/gpteacher-gene... Dataset:vezora/tested-22k-pyth... Endpoints compatible Mistral Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/Lambent/danube2-upscale-1.7

Danube2 Upscale 1.7 Parameters and Internals

Model Type

pre-trained, language model

Additional Notes

Training methodology is experimental, not easy to replicate.

Training Details

Data Sources:

HuggingFaceTB/cosmopedia-100k, Vezora/Tested-22k-Python-Alpaca, sordonia/redpajama-sample_from_valid_all, nampdn-ai/tiny-bridgedict, teknium/GPTeacher-General-Instruct, Severian/Internal-Knowledge-Map, Severian/Internal-Knowledge-Map-StoryWriter-RolePlaying

Methodology:

Linear merge of layers, duplicating layers 16-21, and attempting various repair methods. Merge achieved using EQ-Bench benchmark.

LLM Name	Danube2 Upscale 1.7
Repository 🤗	https://huggingface.co/Lambent/danube2-upscale-1.7
Base Model(s)	Lambent/danube2-upscale-1.53lisa Lambent/danube2-upscale-1.51galore Lambent/danube2-upscale-1.531qlora Lambent/danube2-upscale-1.51qlora Lambent/danube2-upscale-1.53lisa Lambent/danube2-upscale-1.51galore Lambent/danube2-upscale-1.531qlora Lambent/danube2-upscale-1.51qlora
Merged Model	Yes
Model Size	2.2b
Required VRAM	4.5 GB
Updated	2026-07-28
Maintainer	Lambent
Model Type	mistral
Model Files	4.5 GB: 1-of-1
Model Architecture	MistralForCausalLM
License	apache-2.0
Context Length	8192
Model Max Length	8192
Transformers Version	4.36.2
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	32000
Torch Data Type	float16

Rank the Danube2 Upscale 1.7 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 Danube2 Upscale 1.7

Best Alternatives	Context / RAM	Downloads	Likes
... Phi 3 Medium AWQ 4bit Smashed	4K / 7.8 GB	9	0