What are the hardware requirements for Zephyr Orpo 141B A35b V0.1 Bpw2.25?

Zephyr Orpo 141B A35b V0.1 Bpw2.25 requires approximately 40 GB of VRAM and supports a context window of 64K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Zephyr Orpo 141B A35b V0.1 Bpw2.25 and how large is it?

Zephyr Orpo 141B A35b V0.1 Bpw2.25 is developed by blockblockblock, a model with 141b 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 Zephyr Orpo 141B A35b V0.1 Bpw2.25?

Zephyr Orpo 141B A35b V0.1 Bpw2.25 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Zephyr Orpo 141B A35b V0.1 Bpw2.25 by blockblockblock — VRAM 40GB, 64K context

Name: Zephyr Orpo 141B A35b V0.1 Bpw2.25
Author: blockblockblock

Zephyr Orpo 141B A35b V0.1 Bpw2.25 is an open-source language model by blockblockblock. Features: 141b LLM, VRAM: 40GB, Context: 64K, License: apache-2.0, MoE, LLM Explorer Score: 0.11.

Arxiv:2311.07911 Arxiv:2403.07691 Base model:mistral-community/m... Base model:quantized:mistral-c... Conversational Dataset:argilla/distilabel-cap... Endpoints compatible Exl2 Generated from trainer Mixtral Moe Orpo Region:us Safetensors Sharded Tensorflow Trl

Model Card on HF 🤗: https://huggingface.co/blockblockblock/zephyr-orpo-141b-A35b-v0.1-bpw2.25

Zephyr Orpo 141B A35b V0.1 Bpw2.25 Parameters and Internals

Model Type

Mixture of Experts (MoE)

Use Cases

Areas:

research, commercial applications

Primary Use Cases:

chat, code, math, reasoning

Limitations:

The model can produce problematic outputs, especially when prompted to do so., Unknown size and composition of the corpus used to train the base model.

Supported Languages

English (primary)

Training Details

Data Sources:

argilla/distilabel-capybara-dpo-7k-binarized

Data Volume:

7k instances

Methodology:

Odds Ratio Preference Optimization (ORPO)

Training Time:

1.3 hours

Hardware Used:

4 nodes of 8 x H100s

Model Architecture:

Mixture of Experts (MoE) with 141B total parameters and 35B active parameters

LLM Name	Zephyr Orpo 141B A35b V0.1 Bpw2.25
Repository 🤗	https://huggingface.co/blockblockblock/zephyr-orpo-141b-A35b-v0.1-bpw2.25
Base Model(s)	mistral-community/Mixtral-8x22B-v0.1 mistral-community/Mixtral-8x22B-v0.1
Model Size	141b
Required VRAM	40 GB
Updated	2026-08-02
Maintainer	blockblockblock
Model Type	mixtral
Model Files	8.6 GB: 1-of-5 8.6 GB: 2-of-5 8.6 GB: 3-of-5 8.6 GB: 4-of-5 5.6 GB: 5-of-5
Model Architecture	MixtralForCausalLM
License	apache-2.0
Context Length	65536
Model Max Length	65536
Transformers Version	4.39.3
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	32000
Torch Data Type	bfloat16

Rank the Zephyr Orpo 141B A35b V0.1 Bpw2.25 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 Zephyr Orpo 141B A35b V0.1 Bpw2.25

Best Alternatives	Context / RAM	Downloads	Likes
...ephyr Orpo 141B A35b V0.1 GGUF	64K / 4.8 GB	1359	0

Note: green Score (e.g. "73.2") means that the model is better than blockblockblock/zephyr-orpo-141b-A35b-v0.1-bpw2.25.