AI models released in January 2024

10 releases from 9 labs in January 2024, including Code Llama 70B, SFR-Embedding-Mistral and text-embedding-3-small and text-embedding-3-large. Every entry is dated to the lab's official announcement and links to it.

Releases
10
Labs
9
Open weights
7

January 2024

10 releases
  1. MetaCodingOpen weights

    Code Llama 70B

    Code Llama 70B is the largest model in the Code Llama family, released in base, Python and Instruct versions and free for research and commercial use.

    Modalities
    text → text
    Parameters
    70B
    License
    Llama 2 Community License
    At launch
    Open weights
    • Same three versions as the original release: Code Llama, Code Llama - Python and Code Llama - Instruct
    • Free for research and commercial use
    Source: Introducing Code Llama, a state-of-the-art large language model for coding
  2. SalesforceEmbeddingOpen weights

    SFR-Embedding-Mistral

    Text embedding model from Salesforce AI Research built on E5-mistral-7b-instruct and Mistral-7B-v0.1, fine-tuned with multi-task training for retrieval, clustering and related tasks.

    Modalities
    text → embedding
    License
    CC-BY-NC-4.0
    At launch
    Open weights
    • Ranked first on the MTEB benchmark at release, averaging 67.6 across 56 datasets
    • Raised retrieval performance from 56.9 for E5-mistral-7b-instruct to 59.0
    • Released for research use under CC BY-NC 4.0
    Source: SFR-Embedding-Mistral: Enhance Text Retrieval with Transfer Learning
  3. OpenAIEmbeddingProprietary

    text-embedding-3-small and text-embedding-3-large

    OpenAI released two new embedding models, text-embedding-3-small and text-embedding-3-large, succeeding text-embedding-ada-002 and adding a dimensions parameter that shortens embeddings while keeping their concept-representing properties.

    Modalities
    text → embedding
    At launch
    API
    • text-embedding-3-large creates embeddings with up to 3072 dimensions
    • Developers can shorten embeddings natively via the dimensions API parameter
    • Announced alongside gpt-4-0125-preview, the gpt-4-turbo-preview alias, and text-moderation-007
    Source: New embedding models and API updates
  4. MidjourneyImageProprietary

    Niji 6

    Anime and illustration model from the Niji series developed with Spellbrush, following detailed prompts more closely, writing simple text placed in quotes and adding --style raw for less anime styling.

    Modalities
    text, image → image
    At launch
    Consumer app
    • Developed in collaboration with Spellbrush
    • A June 7, 2024 update improved rendering of Japanese kana and simple Chinese characters
    • Supports Character Reference
    Source: Welcome to Niji V6!
  5. 01.AIMultimodalOpen weights

    Yi-VL

    Yi-VL is 01.AI's open bilingual vision-language model family in 6B and 34B sizes, pairing a CLIP ViT-H/14 vision encoder with Yi language models for English and Chinese multimodal chat.

    Modalities
    text, image → text
    Parameters
    6B, 34B
    License
    Apache-2.0
    At launch
    Open weights
    • Image understanding at 448 by 448 resolution
    • Supports English and Chinese conversations, including text recognition in images
    • Trained in three stages covering the ViT, projection module, and full model
    Source: GitHub - 01-ai/Yi: A series of large language models trained from scratch by developers @01-ai
  6. Stability AISmall modelOpen weights

    Stable LM 2 1.6B

    First model of the Stable LM 2 series, a 1.6B-parameter small language model with base and instruction-tuned versions trained on about 2 trillion tokens in seven languages.

    Modalities
    text → text
    Parameters
    1.6B
    License
    Stability AI Community License
    At launch
    Open weights
    • Languages: English, Spanish, German, Italian, French, Portuguese and Dutch
    • Commercial use through a Stability AI Membership
    • Updated version with tool use released April 8, 2024
    Source: Introducing Stable LM 2 1.6B
  7. Alibaba QwenMultimodalProprietary

    Qwen-VL-Max

    Qwen-VL-Max is Qwen's most capable vision-language model at launch, an upgrade over Qwen-VL-Plus with better detailed and text recognition. It supports images up to millions of pixels at arbitrary aspect ratios.

    Modalities
    text, image → text
    At launch
    API, Consumer app
    • Accessible through Hugging Face Spaces, the Qwen website, and DashScope APIs
    • Upgraded alongside Qwen-VL-Plus
    • Proprietary successor to the open Qwen-VL
    Source: Qwen-VL (GitHub)
  8. Stability AICodingOpen weights

    Stable Code 3B

    3B-parameter code completion model built on Stable LM 3B and trained on 18 programming languages, adding fill-in-the-middle support and training on sequences up to 16,384 tokens.

    Modalities
    text → text
    Context
    16K context
    Parameters
    3B
    License
    Stability AI Community License
    At launch
    Open weights
    • Fill-in-the-middle code completion
    • Trained on 18 programming languages
    • Included in the Stability AI Membership for commercial use
    Source: Stable Code 3B: Coding on the Edge
  9. Nous ResearchLLMOpen weights

    Nous-Hermes-2-Mixtral-8x7B

    Hermes 2 fine-tune of Mistral AI's Mixtral 8x7B mixture-of-experts model from Nous Research, released in supervised fine-tuned and DPO-trained versions on Hugging Face.

    Modalities
    text → text
    License
    Apache-2.0
    At launch
    Open weights
    • SFT and DPO versions released together
    • Quantized GGUF version published the same day
    • Built on Mixtral 8x7B
    Source: Releases | Nous Research
  10. DeepSeekLLMOpen weights

    DeepSeekMoE 16B

    DeepSeekMoE 16B is a mixture-of-experts language model that uses fine-grained expert segmentation and shared expert isolation. Trained on 2T English and Chinese tokens, it matches DeepSeek 7B with about 40% of the computation.

    Modalities
    text → text
    Parameters
    MoE: 16.4B total
    License
    DeepSeek Model License
    At launch
    Open weights
    • 16.4B total parameters
    • Deployable on a single 40GB GPU without quantization
    • Released as Base and Chat models
    Source: DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models (GitHub)

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 January 2024." The AI Engineering, updated Jan 29, 2024. https://theaiengineering.com/releases/2024/01

HTML link

<a href="https://theaiengineering.com/releases/2024/01">AI models released in January 2024</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