AI models released in March 2023

10 releases from 8 labs in March 2023, including Dolly, Gen-2 and ERNIE Bot. Every entry is dated to the lab's official announcement and links to it.

Releases
10
Labs
8
Open weights
2

March 2023

10 releases
  1. DatabricksLLMOpen 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
  2. RunwayVideoProprietary

    Gen-2

    Multimodal video generation system that added pure text-to-video to Runway's Gen line, alongside text plus image to video, image to video and video-to-video stylization modes.

    Modalities
    text, image, video → video
    At launch
    Consumer app
    • Announced by Runway as Text to Video with Gen-2
    • Modes include text to video, text plus image to video and image to video
    • Also stylizes existing videos from an image or text prompt
    Source: Runway on X: Introducing, Text to Video. With Gen-2.
  3. BaiduLLMProprietary

    ERNIE Bot

    ERNIE Bot is Baidu's large language model and generative AI product, launched with abilities in literary creation, business writing, mathematical calculation, Chinese language understanding, and multimodal generation.

    Modalities
    text → text, image, audio, video
    At launch
    Consumer app, API
    • Opened to invited users with invitation codes on March 16, 2023
    • Enterprise API access offered on application through Baidu AI Cloud
    • Baidu demonstrated generation of text, images, audio, and video from text prompts
    Source: Baidu Unveils ERNIE Bot, the Latest Generative AI Mastering Chinese Language and Multi-Modal Generation
  4. MidjourneyImageProprietary

    Midjourney V5

    Fifth-generation Midjourney image model with very high coherency, better interpretation of natural language prompts, higher resolution output and support for repeating patterns with --tile.

    Modalities
    text, image → image
    At launch
    Consumer app
    • Default model from March 30, 2023 to May 3, 2023
    • Supported any aspect ratio, up from 2:1 or 1:2 on V4
    • Produced more photographic images than V5.1 but often needed longer prompts
    Source: Legacy Features - Midjourney
  5. AnthropicLLMProprietary

    Claude and Claude Instant

    Claude is Anthropic's first publicly offered AI assistant model, released after a closed alpha with partners such as Notion, Quora, and DuckDuckGo. It launched in two versions: Claude and the lighter, faster, less expensive Claude Instant.

    Modalities
    text → text
    At launch
    API
    • Accessible through a chat interface and the API in Anthropic's developer console
    • Access by request at launch
    • Offered through partner products including Quora's Poe, Notion AI, and DuckDuckGo's DuckAssist
    Source: Introducing Claude
  6. OpenAILLMProprietary

    GPT-4

    GPT-4 is a large multimodal model from OpenAI that accepts image and text inputs and emits text, released with text input in ChatGPT Plus and through a waitlisted API, with image input held back as a research preview.

    Modalities
    text, image → text
    Context
    8K context
    At launch
    Consumer app, API
    • API model gpt-4 (snapshot gpt-4-0314) with an 8,192-token context
    • gpt-4-32k with a 32,768-token context offered in limited access
    • Developers can set the model's style and task through the system message
    • API general availability for developers followed in July 2023
    Source: GPT-4
  7. Z.aiLLMOpen weights

    ChatGLM-6B

    ChatGLM-6B is Z.ai's open bilingual Chinese and English dialogue model with 6.2 billion parameters, based on the GLM architecture. It is designed for local deployment on consumer GPUs.

    Modalities
    text → text
    Context
    2K context
    Parameters
    6.2B
    License
    ChatGLM-6B License
    At launch
    Open weights
    • 6.2 billion parameters
    • 2K context length, later extended in ChatGLM2-6B
    • Code under Apache 2.0; weights under the ChatGLM-6B License
    Source: ChatGLM-6B (GitHub)
  8. AI21 LabsLLMProprietary

    Jurassic-2

    AI21's second generation of foundation models, offered as Large, Grande and Jumbo base models plus instruction-tuned Grande and Jumbo versions through its API and playground.

    Modalities
    text → text
    At launch
    API
    • Supports Spanish, French, German, Portuguese, Italian and Dutch
    • Announced alongside AI21's task-specific APIs
    • Available in the AI21 Studio playground and API
    Source: Announcing Jurassic-2 and Task-Specific APIs
  9. OpenAILLMProprietary

    GPT-3.5 Turbo (gpt-3.5-turbo)

    gpt-3.5-turbo brought the model family behind ChatGPT to the API through a new Chat Completions endpoint that takes a sequence of messages, and OpenAI positioned it for many non-chat use cases as well.

    Modalities
    text → text
    At launch
    API
    • Same model family used in the ChatGPT product
    • Pinned snapshot gpt-3.5-turbo-0301 released the same day
    • Messages are rendered to the model in a format called Chat Markup Language (ChatML)
    • Dedicated instances offered for customers who want reserved capacity
    Source: Introducing ChatGPT and Whisper APIs
  10. OpenAISpeechProprietary

    Whisper API (whisper-1)

    OpenAI made its open-source Whisper large-v2 speech recognition model available as a hosted API model, whisper-1, through transcription and translation endpoints backed by an optimized serving stack.

    Modalities
    audio → text
    At launch
    API
    • Transcriptions endpoint keeps the source language; translations endpoint outputs English
    • Accepts m4a, mp3, mp4, mpeg, mpga, wav, and webm files
    • Serves Whisper large-v2; OpenAI open-sourced Whisper in September 2022
    • Declared generally available in July 2023 alongside the GPT-4 API
    Source: Introducing ChatGPT and Whisper APIs

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. "AI models released in March 2023." The AI Engineering, updated Mar 24, 2023. https://theaiengineering.com/releases/2023/03

HTML link

<a href="https://theaiengineering.com/releases/2023/03">AI models released in March 2023</a> (The AI Engineering)
Full dataset
1,167 releases from 55 labs
Coverage
to
Updated
License
CC BY 4.0
Publisher
The AI Engineering

Archive