What are the hardware requirements for Magnum 32B V2?

Magnum 32B V2 requires approximately 64.6 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 Magnum 32B V2 and how large is it?

Magnum 32B V2 is developed by anthracite-org, a model with 32b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Magnum 32B V2?

Magnum 32B V2 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Magnum 32B V2 by anthracite-org — VRAM 64.6GB, 32K context

Name: Magnum 32B V2
Author: anthracite-org

Magnum 32B V2 is an open-source language model by anthracite-org. Features: 32b LLM, VRAM: 64.6GB, Context: 32K, License: other, LLM Explorer Score: 0.13.

Autotrain compatible Chat Conversational En Endpoints compatible Qwen2 Region:us Safetensors Sharded Tensorflow Zh

Model Card on HF 🤗: https://huggingface.co/anthracite-org/magnum-32b-v2

Magnum 32B V2 Parameters and Internals

Model Type

text-generation

Additional Notes

This is the third model in a series meant to replicate the prose quality of the Claude 3 models, specifically Sonnet and Opus.

Supported Languages

en (English), zh (Chinese)

Training Details

Data Sources:

Stheno dataset (filtered), NobodyExistsOnTheInternet/claude_3.5s_single_turn_unslop_filtered, NobodyExistsOnTheInternet/PhiloGlanSharegpt, NobodyExistsOnTheInternet/Magpie-Reasoning-Medium-Subset, kalomaze/Opus_Instruct_25k, Nopm/Opus_WritingStruct, Gryphe/Sonnet3.5-SlimOrcaDedupCleaned

Methodology:

Instruct tuned with the ChatML formatting

Training Time:

2 epochs

Hardware Used:

8x NVIDIA H100 Tensor Core GPUs

Input Output

Input Format:

ChatML formatting.

LLM Name	Magnum 32B V2
Repository 🤗	https://huggingface.co/anthracite-org/magnum-32b-v2
Model Size	32b
Required VRAM	64.6 GB
Updated	2024-08-18
Maintainer	anthracite-org
Model Type	qwen2
Model Files	4.9 GB: 1-of-14 4.8 GB: 2-of-14 4.8 GB: 3-of-14 4.8 GB: 4-of-14 4.8 GB: 5-of-14 4.8 GB: 6-of-14 4.8 GB: 7-of-14 4.8 GB: 8-of-14 4.8 GB: 9-of-14 4.8 GB: 10-of-14 4.8 GB: 11-of-14 4.8 GB: 12-of-14 4.8 GB: 13-of-14 2.1 GB: 14-of-14 0.0 GB 0.0 GB
Supported Languages	en zh
Model Architecture	Qwen2ForCausalLM
License	other
Context Length	32768
Model Max Length	32768
Transformers Version	4.43.1
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Vocabulary Size	152064
Torch Data Type	bfloat16
Errors	replace

Rank the Magnum 32B V2 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 Magnum 32B V2

Best Alternatives	Context / RAM	Downloads	Likes
Openbuddy Qwq 32B V24.1 200K	195K / 65.8 GB	10	3
Openbuddy Qwq 32B V24.2 200K	195K / 65.8 GB	9	3
Openbuddy Qwq 32B V25.2q 200K	195K / 65.8 GB	1	4
...y Qwen2.5coder 32B V24.1q 200K	195K / 65.8 GB	9	2
DeepSeek R1 Distill Qwen 32B	128K / 65.7 GB	841633	1584
Qwen2.5 32B	128K / 65.5 GB	49206	180
RomboUltima 32B	128K / 20.7 GB	12	6
...wen2.5 32B Inst BaseMerge TIES	128K / 65.8 GB	54	17
Ultiima 32B	128K / 65.8 GB	11	7
...wen2.5 32B Inst BaseMerge TIES	128K / 65.8 GB	47	4

Note: green Score (e.g. "73.2") means that the model is better than anthracite-org/magnum-32b-v2.