Model · mpt-7b-chat

MPT-7B-Chat

Developer: MosaicML
Availability: removed · checked 2026-09-24 · source
HF API returned HTTP 401. Card read from an Internet Archive snapshot dated 2023-07-18.

Raw record: /data/models/mpt-7b-chat.json

Fields

id

mpt-7b-chat

identifiers

huggingface

mosaicml/mpt-7b-chat

developer

MosaicML

release_date

2023-05-05recorded · source

note:  Release post (MosaicML/Databricks blog) and the archived card ('Model Date May 5, 2023'). HF API returned 401 for this repo (mosaicml/ org unreachable, same as mpt-7b).

weights_status

unknown

availability

removed · checked 2026-09-24 · source

note:  HF API returned HTTP 401. Card read from an Internet Archive snapshot dated 2023-07-18.

license

CC-By-NC-SA-4.0recorded · source

note:  Blog: 'CC-By-NC-SA-4.0 (non-commercial use only)'; the archived card says the same. Differs from mpt-7b's Apache-2.0.

architecture

family

decoder_only

n_layers

32recorded · source

note:  Archived card architecture table: n_layers 32, n_heads 32, d_model 4096, vocab size 50432, sequence length 2048.

hidden_size

4096recorded · source

n_heads

32recorded · source

vocab_size

50432recorded · source

context_length

2048recorded · source

positional_encoding

ALiBi (no position embeddings)recorded · source

note:  Blog: MPT 'replaces positional embeddings with ALiBi'. Same as mpt-7b-instruct's record.

note

Archived card: 'follows a modified decoder-only transformer architecture'.

training_data

MPT-7B finetuned on the ShareGPT-Vicuna, HC3, Alpaca, HH-RLHF and Evol-Instruct datasets.recorded · source

note:  Archived card: 'It was built by finetuning MPT-7B on the ShareGPT-Vicuna, HC3, Alpaca, HH-RLHF, and Evol-Instruct datasets.' Training run: 8.2 hours on 8 A100-80GB, then 6.7 hours on 32 A100-40GB.

techniques

primary_sources

https://huggingface.co/mosaicml/mpt-7b-chat

http://web.archive.org/web/20230718105903/https://huggingface.co/mosaicml/mpt-7b-chat

https://www.databricks.com/blog/mpt-7b

record_history

date:2026-09-24 · change:ingested as candidate from HF (mosaicml/mpt-7b-chat@None) · by:ingest_hf.py ·

date:2026-09-24 · change:preparer: developer, release date, license, architecture (from archived card), training data, 3 edges prepared; sources: archived card 2023-07-18, MosaicML/Databricks blog · by:claude (preparer, Sonnet 5) ·

date:2026-09-24 · change:reviewed and promoted from staging (3 edge(s) accepted) · by:Wilson Pruitt ·

Parents

Weights descend

• fine_tuned_from → MPT-7B declared source Blog: MPT-7B-Chat 'built by finetuning MPT-7B'; archived card: 'built by finetuning MPT-7B on ...'.

Training data

• trained_on → ShareGPT conversations (LMSYS collection) declared source Archived card names 'ShareGPT-Vicuna' (link jeffwan/sharegpt_vicuna, unreachable). The `sharegpt-vicuna` record is the LMSYS collection; mapping by name. Reviewer: confirm the id mapping.
• trained_on → Alpaca instruction data (52K) declared source Archived card lists Alpaca (link tatsu-lab/alpaca).

Children

No edges recorded.

Read in

No station on the reading path has touched this record yet.