What are the hardware requirements for BigWeave V15 103B?

BigWeave V15 103B requires approximately 206.6 GB of VRAM and supports a context window of 31K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed BigWeave V15 103B and how large is it?

BigWeave V15 103B is developed by llmixer, a model with 103b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does BigWeave V15 103B perform on standard benchmarks?

BigWeave V15 103B has the following published scores: MMLU 71.25. Compare against reference models on this page or on the LLM Explorer leaderboards.

BigWeave V15 103B by llmixer — VRAM 206.6GB, 31K context

Name: BigWeave V15 103B
Author: llmixer

BigWeave V15 103B is an open-source language model by llmixer. Features: 103b LLM, VRAM: 206.6GB, Context: 31K, License: unknown, LLM Explorer Score: 0.12, Arc: 69.7, HellaSwag: 86.4, MMLU: 71.3, GSM8K: 56.2.

103b Conversational En Endpoints compatible Frankenmerge Llama Model-index Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/llmixer/BigWeave-v15-103b

BigWeave V15 103B Parameters and Internals

Model Type

text-generation

Use Cases

Areas:

research, commercial applications

Additional Notes

The version number tracks attempts, not a quality indicator. Only results with good performance are shared.

Training Details

Methodology:

Self-merge of 152334H/miqu-1-70b-sf, exl2 measurements to identify relevant layers which are then duplicated in pairs.

Input Output

Input Format:

Prompting Format: Mistral, Vicuna and Alpaca.

LLM Name	BigWeave V15 103B
Repository 🤗	https://huggingface.co/llmixer/BigWeave-v15-103b
Model Size	103b
Required VRAM	206.6 GB
Updated	2026-07-12
Maintainer	llmixer
Model Type	llama
Model Files	9.6 GB: 1-of-22 9.8 GB: 2-of-22 9.8 GB: 3-of-22 9.9 GB: 4-of-22 9.8 GB: 5-of-22 10.0 GB: 6-of-22 9.8 GB: 7-of-22 10.0 GB: 8-of-22 9.8 GB: 9-of-22 9.8 GB: 10-of-22 9.8 GB: 11-of-22 9.8 GB: 12-of-22 9.8 GB: 13-of-22 10.0 GB: 14-of-22 9.7 GB: 15-of-22 9.9 GB: 16-of-22 9.6 GB: 17-of-22 10.0 GB: 18-of-22 9.8 GB: 19-of-22 9.7 GB: 20-of-22 9.9 GB: 21-of-22 0.3 GB: 22-of-22
Supported Languages	en
Model Architecture	LlamaForCausalLM
License	unknown
Context Length	32764
Model Max Length	32764
Transformers Version	4.36.2
Tokenizer Class	LlamaTokenizer
Padding Token	<unk>
Vocabulary Size	32000
Torch Data Type	float16

Rank the BigWeave V15 103B 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 BigWeave V15 103B

Best Alternatives	Context / RAM	Downloads	Likes
Dark Miqu 103B	31K / 206.4 GB	30	2
Midnight Miqu 103B V1.5	31K / 206.4 GB	33	28
Midnight Miqu 103B V1.0	31K / 206.4 GB	85	16
Midnight Rose 103B V2.0.3	4K / 206.6 GB	93	13
OmegaDolphin 103B V0.1	4K / 175.6 GB	6	1
KitchenSink 103b	4K / 178.6 GB	13	9
Midnight Rose 103B V1.0	4K / 206.6 GB	11	3
Aurora Nights 103B V1.0	4K / 206.6 GB	6	14
Rogue Rose 103B V0.2	4K / 206.5 GB	12	33
Lila 103B L2	4K / 206.2 GB	0	0

Note: green Score (e.g. "73.2") means that the model is better than llmixer/BigWeave-v15-103b.