Databricks models

Data and AI platform company that released the Dolly instruction-tuned models in 2023, acquired MPT maker MosaicML that July and released the DBRX mixture-of-experts model in 2024.

San Francisco, United StatesFounded 2013databricks.com

Releases
5
Open weights
5
Latest
Mar 27, 2024

March 2024

1 release
  1. LLMOpen weightsRetired

    DBRX

    Open general-purpose language model from Databricks built on a fine-grained mixture-of-experts architecture, released as DBRX Base and DBRX Instruct for download and API use.

    Modalities
    text → text
    Context
    32K context
    Parameters
    MoE: 132B total, 36B active
    License
    Databricks Open Model License
    At launch
    Open weights, API
    Retired
    • Pretrained on 12 trillion tokens of curated data
    • Inference up to twice as fast as LLaMA2-70B
    • Offered to Databricks customers through APIs at launch

    Databricks retired DBRX and DBRX Instruct from pay-per-token serving on April 30, 2025 and from provisioned throughput on December 19, 2025.

    Source: Introducing DBRX: A New State-of-the-Art Open LLM | Databricks Blog

    Retirement: AI models maintenance policy | Databricks on AWS

June 2023

1 release
  1. LLMOpen weightsRetired

    MPT-30B

    MosaicML Foundation Series model trained with an 8k-token context window, released by MosaicML four days before Databricks announced its agreement to acquire the company.

    Modalities
    text → text
    Context
    8K context
    Parameters
    30B
    License
    Apache-2.0
    At launch
    Open weights, API
    Retired
    • Pretrained on 1T tokens at 2k context, then 50B tokens at 8k
    • Partly trained on NVIDIA H100 GPUs
    • Released with Instruct and non-commercial Chat variants

    Databricks retired MPT-30B and MPT-30B Instruct from pay-per-token serving on August 30, 2024 and from provisioned throughput on December 19, 2025.

    Source: MPT-30B: Raising the bar for open-source foundation models | Databricks Blog

    Retirement: AI models maintenance policy | Databricks on AWS

May 2023

1 release
  1. LLMOpen weightsRetired

    MPT-7B

    First MosaicML Foundation Series model, a decoder-only transformer trained from scratch on 1T tokens of text and code and licensed for commercial use, released by MosaicML before Databricks acquired the company in July 2023.

    Modalities
    text → text
    Parameters
    6.7B
    License
    Apache-2.0
    At launch
    Open weights
    Retired
    • Trained on the MosaicML platform in 9.5 days for about $200k
    • StoryWriter-65k+ variant fine-tuned with a 65k-token context
    • Instruct variant under CC BY-SA 3.0, Chat variant non-commercial

    Databricks retired MPT-7B and MPT-7B Instruct from pay-per-token serving on August 30, 2024 and from provisioned throughput on December 19, 2025.

    Source: Introducing MPT-7B: A New Standard for Open-Source, Commercially Usable LLMs | Databricks Blog

    Retirement: AI models maintenance policy | Databricks on AWS

April 2023

1 release
  1. LLMOpen weights

    Dolly 2.0 (dolly-v2-12b)

    12B instruction-following model based on EleutherAI's Pythia family, fine-tuned only on databricks-dolly-15k, a human-written instruction dataset licensed for research and commercial use.

    Modalities
    text → text
    Parameters
    12B
    License
    MIT
    At launch
    Open weights
    • Dataset written by more than 5,000 Databricks employees in March and April 2023
    • Training code, dataset and weights released for commercial use
    • databricks-dolly-15k licensed under CC BY-SA 3.0
    Source: Dolly: Open Instruction-Tuned LLM | Databricks Blog

March 2023

1 release
  1. LLMOpen weights

    Dolly

    Databricks' first instruction-following model, made by fine-tuning a two-year-old 6B open model from EleutherAI on Alpaca data to show that instruction following does not seem to require the latest or largest models.

    Modalities
    text → text
    Parameters
    6B
    License
    CC-BY-NC-4.0
    At launch
    Open weights
    • Trained in 30 minutes on one machine
    • Fine-tuned on instruction data from Stanford's Alpaca
    • Released with a Databricks notebook for rebuilding it
    Source: Hello Dolly: Democratizing the magic of ChatGPT with open models

Cite and reuse this data

These release records are free to reuse under CC BY 4.0, including commercially. Credit The AI Engineering and link to this page's URL.

Dates are the labs' official announcement dates, and every record links to its source. List fields in the CSV are separated by " | ".

Citation

The AI Engineering Editors. "Databricks models: release history and timeline." The AI Engineering, updated Mar 27, 2024. https://theaiengineering.com/releases/databricks

HTML link

<a href="https://theaiengineering.com/releases/databricks">Databricks models: release history and timeline</a> (The AI Engineering)
Full dataset
1,167 releases from 55 labs
Coverage
to
Updated
License
CC BY 4.0
Publisher
The AI Engineering

Other labs