What are the hardware requirements for Lumina 3?

Lumina 3 requires approximately 37.1 GB of VRAM and supports a context window of 32K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Lumina 3 and how large is it?

Lumina 3 is developed by Ppoyaa, a model with 7b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Lumina 3 perform on standard benchmarks?

Lumina 3 has the following published scores: MMLU 65.52. Compare against reference models on this page or on the LLM Explorer leaderboards.

Lumina 3 by Ppoyaa — VRAM 37.1GB, 32K context

Name: Lumina 3
Author: Ppoyaa

Lumina 3 is an open-source language model by Ppoyaa. Features: 7b LLM, VRAM: 37.1GB, Context: 32K, License: apache-2.0, LLM Explorer Score: 0.13, Arc: 71.2, HellaSwag: 87.2, MMLU: 65.5, GSM8K: 72.9.

Conversational Endpoints compatible Frankenmoe Lazymergekit Merge Mergekit Mixtral Moe Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/Ppoyaa/Lumina-3

Lumina 3 Parameters and Internals

Model Type

Mixture of Experts (MoE)

Supported Languages

language (en), proficiency (High)

Training Details

Methodology:

LazyMergekit

Context Length:

32000

LLM Name	Lumina 3
Repository 🤗	https://huggingface.co/Ppoyaa/Lumina-3
Model Size	7b
Required VRAM	37.1 GB
Updated	2026-05-28
Maintainer	Ppoyaa
Model Type	mixtral
Model Files	1.9 GB: 1-of-19 2.0 GB: 2-of-19 2.0 GB: 3-of-19 2.0 GB: 4-of-19 2.0 GB: 5-of-19 2.0 GB: 6-of-19 2.0 GB: 7-of-19 2.0 GB: 8-of-19 2.0 GB: 9-of-19 2.0 GB: 10-of-19 2.0 GB: 11-of-19 2.0 GB: 12-of-19 2.0 GB: 13-of-19 2.0 GB: 14-of-19 2.0 GB: 15-of-19 2.0 GB: 16-of-19 2.0 GB: 17-of-19 2.0 GB: 18-of-19 1.2 GB: 19-of-19
Model Architecture	MixtralForCausalLM
License	apache-2.0
Context Length	32768
Model Max Length	32768
Transformers Version	4.39.3
Tokenizer Class	LlamaTokenizer
Padding Token	<s>
Vocabulary Size	32000
Torch Data Type	bfloat16

Rank the Lumina 3 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 Lumina 3

Best Alternatives	Context / RAM	Downloads	Likes
Multimaster 7B V6	32K / 142.5 GB	8108	2
Mini Mixtral V0.2	32K / 25.8 GB	3923	4
MultiverseBuddy 15B MoE	32K / 25.8 GB	9	0
Lumina 2	32K / 37.1 GB	6	0
OpenMistral MoE	32K / 48.3 GB	1218	0
RogerWizard 12B MoE	32K / 25.8 GB	4	1
Rawr	32K / 93.5 GB	1302	0
Merged Model MoE	32K / 53.3 GB	7	1
StarlingMaths 12B MoE	32K / 25.8 GB	5	0
MultiverseMath 12B MoE	32K / 25.8 GB	6	0

Note: green Score (e.g. "73.2") means that the model is better than Ppoyaa/Lumina-3.