Name: Mpt 30B Instruct
Author: mosaicml

Mpt 30B Instruct is an open-source language model by mosaicml. Features: 30b LLM, VRAM: 60.1GB, License: apache-2.0, Instruction-Based, HF Score: 53.4, LLM Explorer Score: 0.13, Arc: 58.5, HellaSwag: 84.3, MMLU: 49.2, TruthfulQA: 38.1, WinoGrande: 75.1, GSM8K: 15.3.

Model Card on HF 🤗: https://huggingface.co/mosaicml/mpt-30b-instruct

Mpt 30B Instruct Benchmarks

ARC: 58.45 vs 96.7 (so35)^-39.6%

HellaSwag: 84.31 vs 95.3 (gpt4)^-11.5%

MMLU: 49.15 vs 88.3 (so35)^-44.3%

TruthfulQA: 38.05 vs 59 (gpt4)^-35.5%

WinoGrande: 75.14 vs 87.5 (gpt4)^-14.1%

GSM8K: 15.31 vs 96.4 (so35)^-84.1%

LLME Score: 0.12877

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

Mpt 30B Instruct (mosaicml/mpt-30b-instruct)

🌟 Advertise your project 🚀

Mpt 30B Instruct Parameters and Internals

Model Type

text generation, instruction following

Use Cases

Areas:

Research, Commercial applications

Primary Use Cases:

Short-form instruction following tasks

Limitations:

May produce factually incorrect output and offensive content

Training Details

Data Sources:

Dolly HHRLHF derived from Databricks Dolly-15k, Anthropic Helpful and Harmless (HH-RLHF), Competition Math, Duorc, CoT GSM8k, Qasper, Quality, Summ Screen FD, Spider

Methodology:

The model was finetuned on short-form instruction following tasks from various datasets.

Context Length:

8192

Training Time:

8 hours on 72 A100 40GB GPUs

Hardware Used:

72 A100 40GB GPUs

Model Architecture:

A modification of a standard decoder-only transformer, utilizes FlashAttention and ALiBi with no positional embeddings.

Input Output

Input Format:

Prompts formatted with instructions preceding responses.

Accepted Modalities:

text

Output Format:

Generated text responses

Performance Tips:

Use torch.autocast for lower precision performance, and `trust_remote_code=True` to load custom architecture.

LLM Name	Mpt 30B Instruct
Repository 🤗	https://huggingface.co/mosaicml/mpt-30b-instruct
Model Size	30b
Required VRAM	60.1 GB
Updated	2025-10-07
Maintainer	mosaicml
Model Type	mpt
Instruction-Based	Yes
Model Files	9.8 GB: 1-of-7 9.9 GB: 2-of-7 9.9 GB: 3-of-7 9.9 GB: 4-of-7 9.9 GB: 5-of-7 9.9 GB: 6-of-7 0.8 GB: 7-of-7
Model Architecture	MPTForCausalLM
License	apache-2.0
Model Max Length	8192
Transformers Version	4.28.1
Tokenizer Class	GPTNeoXTokenizer
Vocabulary Size	50432
Torch Data Type	bfloat16

Quantized Models of the Mpt 30B Instruct

Model	Likes	Downloads	VRAM
Mpt 30B Instruct Q8	5	10	30 GB

Best Alternatives to Mpt 30B Instruct

Best Alternatives	Context / RAM	Downloads	Likes
...t 30B Instruct Peft Compatible	0K / 60.1 GB	0	2
Ct2fast Mpt 30B Instruct	0K / 30 GB	7	4
Ct2fast Mpt 30B Chat	0K / 30 GB	9	2
Mpt 30B Chat Q8	0K / 30.4 GB	94	1
Mpt 30B Instruct Q8	0K / 30.4 GB	10	5
...l Mpt 30B Instruct W4 G128 AWQ	0K / 16.1 GB	5	2

Note: green Score (e.g. "73.2") means that the model is better than mosaicml/mpt-30b-instruct.

Rank the Mpt 30B Instruct Capabilities

🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟

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

What open-source LLMs or SLMs are you in search of? 52392 in total.

Email us: info@extractum.io. Our Privacy Policy | Terms and Conditions | Suggest an improvement.

Our Social Media →

Original data from HuggingFace, OpenCompass and various public git repos.

Check out Ag3ntum — our secure, self-hosted AI agent for server management.

Release v20260328a

Support LLM Explorer

Mpt 30B Instruct by mosaicml

» All LLMs » mosaicml » Mpt 30B Instruct URL Share it on