LLM EXPLORER 59,420 MODELS INDEXED

Mpt 7B Chat Q8 by Abzu

By Abzu · 17 downloads
HF Repo ↗ Copy URL

Mpt 7B Chat Q8 is an open-source language model by Abzu. Features: 7b LLM, VRAM: 6.9GB, License: cc-by-nc-sa-4.0, Quantized, Instruction-Based, LLM Explorer Score: 0.21, ELO: 1063, Arc: 46.5, HellaSwag: 75.5, MMLU: 37.6, GSM8K: 4.1.

  Arxiv:2010.04245   Arxiv:2108.12409   Arxiv:2205.14135   8-bit   Composer   Custom code   Dataset:anthropic/hh-rlhf   Dataset:hello-simpleai/hc3 Dataset:jeffwan/sharegpt vicun...   Dataset:tatsu-lab/alpaca Dataset:victor123/evol instruc...   Instruct   Llm-foundry   Mosaicml   Mpt   Q8   Quantized   Region:us   Safetensors
Model Card on HF 🤗: https://huggingface.co/Abzu/mpt-7b-chat-q8 

Mpt 7B Chat Q8 Benchmarks

LMSys ELO 1063 vs 1272 (so35) -16.4%
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

Mpt 7B Chat Q8 Parameters and Internals

Model Type 
Dialogue generation, Chatbot
Additional Notes 
Trained with sequence length of 2048, using various attention and optimization techniques, able to extend sequence length up to 4096.
Training Details 
Data Sources:
jeffwan/sharegpt_vicuna, Hello-SimpleAI/HC3, tatsu-lab/alpaca, Anthropic/hh-rlhf, victor123/evol_instruct_70k
Methodology:
Finetuning
Context Length:
2048
Hardware Used:
8 A100-80GB GPUs, 32 A100-40GB GPUs
Model Architecture:
Modified decoder-only transformer
LLM NameMpt 7B Chat Q8
Repository 🤗https://huggingface.co/Abzu/mpt-7b-chat-q8 
Base Model(s)  ...uct Gpt4all Max Length 3072 V1   NickyNicky/mpt-7b-chat-Peft-h2ogpt_oig_oasst1_instruct-gpt4all-max_length_3072-V1
Model Size7b
Required VRAM6.9 GB
Updated2026-08-25
MaintainerAbzu
Model Typempt
Instruction-BasedYes
Model Files  6.9 GB
Quantization Typeq8
Model ArchitectureMPTForCausalLM
Licensecc-by-nc-sa-4.0
Model Max Length2048
Transformers Version4.30.2
Tokenizer ClassGPTNeoXTokenizer
Vocabulary Size50432
Torch Data Typebfloat16

Best Alternatives to Mpt 7B Chat Q8

Best Alternatives
Context / RAM
Downloads
Likes
Mpt 7B Instruct Q80K / 6.9 GB182
Mpt 7B Chat0K / 13.3 GB80920518
Mpt 7B Instruct0K / 13.3 GB7946470
Mpt 7B Int8 Ov0K / 0 GB130
Mpt 7B 8K Instruct0K / 13.3 GB201227
SEA LION V1 7B IT GBTQ0K / 5.5 GB70
Sea Lion 7B Instruct Gptq0K / 5.5 GB50
SEA LION V1 7B IT GPTQ0K / 5.5 GB50
Sea Lion 7B Instruct0K / 15 GB20823
Sea Lion 7B Instruct Research0K / 15 GB1114
Note: green Score (e.g. "73.2") means that the model is better than Abzu/mpt-7b-chat-q8.