What are the hardware requirements for Dolphin 2.9.1 Mistral 22B?

Dolphin 2.9.1 Mistral 22B requires approximately 44.7 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 Dolphin 2.9.1 Mistral 22B and how large is it?

Dolphin 2.9.1 Mistral 22B is developed by theo77186, a model with 22.2b 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 Dolphin 2.9.1 Mistral 22B?

Dolphin 2.9.1 Mistral 22B is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Dolphin 2.9.1 Mistral 22B by theo77186 — VRAM 44.7GB, 64K context

Name: Dolphin 2.9.1 Mistral 22B
Author: theo77186

Dolphin 2.9.1 Mistral 22B is an open-source language model by theo77186. Features: 22.2b LLM, VRAM: 44.7GB, Context: 64K, License: apache-2.0, MoE, LLM Explorer Score: 0.12.

Conversational Endpoints compatible Mistral Moe Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/theo77186/dolphin-2.9.1-mistral-22b

Dolphin 2.9.1 Mistral 22B Parameters and Internals

Additional Notes

ChatML functionality is broken. Alpaca works despite not being specifically trained on it.

Training Details

Data Sources:

GPT-4, among other models

Methodology:

Fully fine-tuned, targeting all layers. Extracted expert using SLERP.

Context Length:

64000

Training Time:

27 hours

Hardware Used:

8xH100 provided by Crusoe Cloud

Model Architecture:

Mixtral architecture with a single expert.

LLM Name	Dolphin 2.9.1 Mistral 22B
Repository 🤗	https://huggingface.co/theo77186/dolphin-2.9.1-mistral-22b
Model Size	22.2b
Required VRAM	44.7 GB
Updated	2026-07-22
Maintainer	theo77186
Model Type	mistral
Model Files	4.9 GB: 1-of-9 5.0 GB: 2-of-9 5.0 GB: 3-of-9 4.9 GB: 4-of-9 5.0 GB: 5-of-9 5.0 GB: 6-of-9 4.9 GB: 7-of-9 5.0 GB: 8-of-9 5.0 GB: 9-of-9
Model Architecture	MistralForCausalLM
License	apache-2.0
Context Length	65536
Model Max Length	65536
Transformers Version	4.40.1
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	32002
Torch Data Type	bfloat16

Rank the Dolphin 2.9.1 Mistral 22B 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 Dolphin 2.9.1 Mistral 22B

Best Alternatives	Context / RAM	Downloads	Likes
Mistral Small Instruct 2409	128K / 44.7 GB	32	8
Mistral Small NovusKyver	128K / 44.7 GB	2	4
WizardLM 2 22B RP	64K / 44.7 GB	9	5
...Iteration Kto 300sm 3ep Merged	32K / 44.7 GB	8	0
...ct MistralSmallit Mg DPO Iter2	32K / 44.7 GB	52	0
...lect MistralSmallit MMQA Iter1	32K / 44.7 GB	36	0
...ct MistralSmallit Mg DPO Iter1	32K / 60.8 GB	38	0
...istral Small It MMQA DPO Iter5	32K / 44.7 GB	56	0
...istral Small It MMQA DPO Iter4	32K / 44.7 GB	45	0
...istral Small It MMQA DPO Iter2	32K / 60.8 GB	7	0

Note: green Score (e.g. "73.2") means that the model is better than theo77186/dolphin-2.9.1-mistral-22b.