AI models released in September 2023

7 releases from 7 labs in September 2023, including Amazon Titan Embeddings G1 - Text, Granite (granite.13b) and Mistral 7B. Every entry is dated to the lab's official announcement and links to it.

Releases
7
Labs
7
Open weights
3

September 2023

7 releases
  1. AmazonEmbeddingProprietary

    Amazon Titan Embeddings G1 - Text

    Amazon Titan Embeddings G1 - Text was the first generally available Titan model, a text embedding model for search, personalization, clustering and retrieval-augmented generation, launched with general availability of Amazon Bedrock.

    Modalities
    text → embedding
    Context
    8K context
    At launch
    API
    • Generally available with Amazon Bedrock on September 28, 2023
    • 8K token context window
    • Previewed with select customers after the April 13, 2023 Titan announcement
    Source: Amazon Bedrock Is Now Generally Available
  2. IBMLLMProprietary

    Granite (granite.13b)

    IBM's first watsonx Granite foundation models, including granite.13b.instruct and granite.13b.chat, reached general availability in watsonx.ai for enterprise language and code tasks.

    Modalities
    text → text
    Parameters
    13B
    At launch
    Enterprise
    • Covered by IBM's standard contractual IP protections
    • Delivered through the watsonx.ai platform
    • First models in the watsonx Granite series
    Source: IBM Announces Availability of watsonx Granite Model Series, Client Protections for IBM watsonx Models
  3. Mistral AILLMOpen weights

    Mistral 7B

    Mistral AI's first model, a 7.3B-parameter open-weights language model released under Apache 2.0 with a companion chat fine-tune, built for efficient inference.

    Modalities
    text → text
    Context
    8K context
    Parameters
    7.3B
    License
    Apache-2.0
    At launch
    Open weights
    • Uses grouped-query attention and sliding window attention
    • Released with a reference implementation and an instruction-tuned chat variant
    • Apache 2.0 license allows use without restrictions
    Source: Mistral 7B | Mistral AI
  4. OpenAIMultimodalProprietary

    GPT-4 with vision (GPT-4V)

    OpenAI began rolling out image input in ChatGPT, powered by multimodal GPT-3.5 and GPT-4 (GPT-4V), letting users discuss photos, screenshots, and documents, alongside new voice conversations on mobile.

    Modalities
    text, image → text
    At launch
    Consumer app, Enterprise
    • Rolled out to ChatGPT Plus and Enterprise users over two weeks
    • Voice chat used a new text-to-speech model plus Whisper for transcription
    • OpenAI published the GPT-4V(ision) system card the same day
    • Vision in the API followed at DevDay via gpt-4-vision-preview
    Source: ChatGPT can now see, hear, and speak
  5. Alibaba QwenLLMOpen weights

    Qwen-14B

    Qwen-14B is a first-generation Qwen base model released with the Qwen-14B-Chat assistant model, along with the qwen.cpp runtime and the Qwen-Agent framework.

    Modalities
    text → text
    Context
    8K context
    Parameters
    14B
    License
    Tongyi Qianwen License Agreement
    At launch
    Open weights
    • Pre-trained on 3.0T tokens
    • 8K maximum length
    • Released on ModelScope and Hugging Face
    Source: Qwen (GitHub)
  6. Stability AIMusicProprietary

    Stable Audio

    Stability AI's first music and sound generation product, generating tracks from text prompts and a requested duration through a web interface with free and Pro tiers.

    Modalities
    text → audio
    At launch
    Consumer app
    • Free tier generates tracks up to 45 seconds
    • Pro subscription produces 90-second tracks for commercial use
    • Users set a desired track length along with the text prompt
    Source: Announcing Stable Audio, a product for music & sound generation
  7. MicrosoftSmall modelOpen weights

    phi-1.5

    phi-1.5 is a 1.3B Microsoft Research language model trained on the phi-1 data sources plus synthetic NLP texts, extending the textbook-quality data approach from Python coding to common sense reasoning and general language.

    Modalities
    text → text
    Parameters
    1.3B
    License
    MIT
    At launch
    Open weights
    • 1.3B parameters trained on 150B tokens
    • Excludes generic web-crawl sources such as Common Crawl
    • Released on Hugging Face under the MIT license
    Source: microsoft/phi-1_5 on Hugging Face

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 September 2023." The AI Engineering, updated Sep 28, 2023. https://theaiengineering.com/releases/2023/09

HTML link

<a href="https://theaiengineering.com/releases/2023/09">AI models released in September 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