AI models released in June 2023

9 releases from 8 labs in June 2023, including XGen-7B, ChatGLM2-6B and Inflection-1. Every entry is dated to the lab's official announcement and links to it.

Releases
9
Labs
8
Open weights
5

June 2023

9 releases
  1. SalesforceLLMOpen weights

    XGen-7B

    Series of 7B language models trained with standard dense attention on sequences up to 8K tokens for up to 1.5 trillion tokens, plus instruction-tuned versions.

    Modalities
    text → text
    Context
    8K context
    Parameters
    7B
    License
    Apache-2.0
    At launch
    Open weights
    • Base models trained at 4K and 8K sequence lengths
    • Instruct versions tuned on public-domain instructional data
    • Base models released under Apache 2.0
    Source: Long Sequence Modeling with XGen: A 7B LLM Trained on 8K Input Sequence Length - Salesforce
  2. Z.aiLLMOpen weights

    ChatGLM2-6B

    ChatGLM2-6B is the second-generation version of Z.ai's open bilingual chat model, with an upgraded base model pre-trained on 1.4T Chinese and English tokens. It extends base context from 2K to 32K tokens and is trained for 8K dialogue.

    Modalities
    text → text
    Context
    32K context
    Parameters
    6B
    License
    ChatGLM2-6B License
    At launch
    Open weights
    • Uses Multi-Query Attention for faster inference and lower memory use
    • Context length extended from 2K to 32K using FlashAttention
    • Free commercial use allowed after registration
    Source: ChatGLM-6B (GitHub)
  3. Inflection AILLMProprietary

    Inflection-1

    Inflection AI's in-house large language model powering Pi, trained on thousands of NVIDIA H100 GPUs and adapted through a proprietary process into a personal AI, with a technical memo detailing its evaluations.

    Modalities
    text → text
    At launch
    Consumer app
    • Powers Pi at pi.ai; Inflection said it would come to its conversational API
    • Technical memo reports results for the base model before fine-tuning or alignment
    Source: Inflection-1: Pi’s Best-in-Class LLM
  4. MidjourneyImageProprietary

    Midjourney V5.2

    Version 5 update producing more detailed, sharper results with better colors, contrast and composition, a slightly better understanding of prompts and stronger response to the --stylize parameter.

    Modalities
    text, image → image
    At launch
    Consumer app
    • Default model from June 22, 2023 to February 14, 2024
    • The Discord Style Tuner was a Version 5.2 feature
    • Midjourney's documentation lists its release as June 2023
    Source: Legacy Features - Midjourney
  5. DatabricksLLMOpen 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

  6. Stability AIImageOpen weights

    SDXL 0.9

    Research release of Stable Diffusion XL with a 3.5B-parameter base model in a 6.6B-parameter two-model ensemble, offered on Clipdrop and the Stability API ahead of the open SDXL 1.0 release.

    Modalities
    text, image → image
    Parameters
    3.5B base, 6.6B ensemble
    License
    SDXL 0.9 Research License
    At launch
    Open weights, API, Consumer app
    • Research weights available on application under a non-commercial, research-only license
    • Uses two CLIP models including OpenCLIP ViT-G/14
    • Native 1024x1024 resolution
    Source: Stability AI launches SDXL 0.9: A Leap Forward in AI Image Generation
  7. MicrosoftCodingOpen weights

    phi-1

    phi-1 is a 1.3B Microsoft Research model for Python code generation, trained on filtered code data and synthetic textbooks and exercises, introduced in the paper Textbooks Are All You Need.

    Modalities
    text → text
    Parameters
    1.3B
    License
    MIT
    At launch
    Open weights
    • 1.3B parameter Transformer specialized in Python
    • Trained on 54B tokens including synthetic textbooks generated with GPT-3.5
    • Weights published on Hugging Face under the MIT license
    Source: microsoft/phi-1 on Hugging Face
  8. OpenAILLMProprietary

    gpt-3.5-turbo-0613 and gpt-3.5-turbo-16k

    OpenAI released updated GPT-3.5 Turbo models: gpt-3.5-turbo-0613 adds function calling and more reliable steerability via the system message, and gpt-3.5-turbo-16k offers four times the context length of the standard 4K version.

    Modalities
    text → text
    At launch
    API
    • gpt-3.5-turbo-16k fits about 20 pages of text in a single request
    • 16,385-token context for gpt-3.5-turbo-16k-0613 per its docs model page
    • Stable gpt-3.5-turbo alias was set to upgrade to the new model on June 27, 2023
    Source: Function calling and other API updates
  9. OpenAILLMProprietary

    gpt-4-0613

    gpt-4-0613 is an updated GPT-4 snapshot that introduced function calling in the Chat Completions API, letting the model output JSON arguments for developer-described functions, alongside gpt-4-32k-0613 with the same improvements and extended context.

    Modalities
    text → text
    Context
    8K context
    At launch
    API
    • 8,192-token context for gpt-4 snapshots per the GPT-4 docs model page
    • gpt-4-32k-0613 adds the same improvements with a longer context
    • Stable gpt-4 and gpt-4-32k aliases were set to upgrade to the new models on June 27, 2023
    Source: Function calling and other API updates

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 June 2023." The AI Engineering, updated Jun 28, 2023. https://theaiengineering.com/releases/2023/06

HTML link

<a href="https://theaiengineering.com/releases/2023/06">AI models released in June 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