AI models released in June 2024
17 releases from 14 labs in June 2024, including DeepSeek-V2-Chat-0628, Gemma 2 and Doubao Speech Synthesis and Voice Cloning. Every entry is dated to the lab's official announcement and links to it.
- Releases
- 17
- Labs
- 14
- Open weights
- 10
June 2024
17 releasesDeepSeek-V2-Chat-0628
DeepSeek-V2-Chat-0628 is an upgraded DeepSeek-V2-Chat whose base model was replaced with the Coder-V2 base, improving code generation, math, and reasoning. It also strengthened role-playing.
- Modalities
- text → text
- License
- DeepSeek Model License
- At launch
- API, Open weights
- Upgraded the deepseek-chat API model on June 28, 2024
- Weights published on Hugging Face as DeepSeek-V2-Chat-0628
- Later merged with DeepSeek-Coder-V2-0724 to form DeepSeek-V2.5
Gemma 2
Gemma 2 is Google's second generation of open models, released in 9B and 27B sizes with a redesigned architecture for performance and inference efficiency. The 27B model can run full-precision inference on a single NVIDIA H100, A100 80GB or TPU host.
- Modalities
- text → text
- Parameters
- 9B, 27B
- License
- Gemma Terms of Use
- At launch
- Open weights, API, Cloud platforms
- Redesigned architecture aimed at performance and inference efficiency
- 27B available to try in Google AI Studio
- Weights on Kaggle and Hugging Face under a commercially friendly Gemma license
Doubao Speech Synthesis and Voice Cloning
Volcano Engine released Doubao's large-model speech synthesis and voice cloning models, which infer emotion and intonation from context to produce natural speech and create custom voices from seconds of recorded audio.
- Modalities
- text, audio → audio
- At launch
- API
- Voice cloning builds a custom voice from seconds of audio
- Speech synthesis predicts emotion and tone from surrounding text
- Offered as APIs on Volcano Engine
Claude 3.5 Sonnet
Claude 3.5 Sonnet is the first release in Anthropic's Claude 3.5 family, running at twice the speed of Claude 3 Opus with stronger coding and vision. It debuted alongside Artifacts, a Claude.ai feature for working with generated content.
- Modalities
- text, image → text
- Context
- 200K context
- At launch
- API, Consumer app, Cloud platforms
- API model ID claude-3-5-sonnet-20240620
- 200K-token context window
- Free on Claude.ai and the Claude iOS app; also on the Anthropic API, Amazon Bedrock, and Vertex AI
Chameleon
Chameleon is a mixed-modal early-fusion model family from Meta FAIR; Meta released key components of the 7B and 34B models for research with text-only output.
- Modalities
- text, image → text
- Parameters
- 7B, 34B
- License
- Chameleon Research License
- At launch
- Open weights
- Accepts mixed text and image input
- Released under a research-only license
- The Chameleon image generation model was not released
Gen-3 Alpha
First of Runway's next-generation models trained on new large-scale multimodal infrastructure, trained jointly on videos and images with major gains in fidelity, consistency and motion over Gen-2.
- Modalities
- text, image → video
- At launch
- Consumer app
- Text to video became available on July 1, 2024
- Image to video added July 29, 2024
- Powers Motion Brush, Advanced Camera Controls and Director Mode
DeepSeek-Coder-V2
DeepSeek-Coder-V2 is an open-weights mixture-of-experts code model further pre-trained from an intermediate DeepSeek-V2 checkpoint with an additional 6T tokens. It substantially improved coding and math reasoning while keeping general language performance.
- Modalities
- text → text
- Context
- 125K context
- Parameters
- MoE: 236B total, 21B active; Lite 16B total, 2.4B active
- License
- DeepSeek Model License
- At launch
- API, Open weights
- Released in 236B and 16B Lite sizes, each with Base and Instruct models
- 128K context length
- Upgraded the deepseek-coder API model on June 14, 2024
Nemotron-4 340B
Family of open base, instruct and reward models released as a pipeline for generating synthetic data to train other language models, under NVIDIA's open model license.
- Modalities
- text → text
- Parameters
- 340B
- License
- NVIDIA Open Model License
- At launch
- Open weights
- Released as Base, Instruct and Reward models
- Downloadable from NVIDIA NGC and Hugging Face
- Instruct model trained between December 2023 and May 2024
Dream Machine
Luma's first video generation model, a multimodal transformer trained directly on videos that makes realistic and fantastical clips from text instructions and images, released free to everyone.
- Modalities
- text, image → video
- At launch
- Consumer app, API
- Built with AWS SageMaker HyperPod H100 infrastructure
- Text-to-video and image-to-video
- Dream Machine API launched in September 2024
Phoenix (preview)
Leonardo.Ai's first foundational image model, built by its own R&D team and launched in preview for all users with a focus on prompt adherence and accurate text in images.
- Modalities
- text → image
- At launch
- Consumer app, API
- Launched alongside Prompt Enhance for shorter prompts
- Available through the Leonardo API from August 30, 2024
- Built by Leonardo's own R&D team
Stable Diffusion 3 Medium
First open-weight release from the Stable Diffusion 3 series, a 2B-parameter text-to-image model, also offered through the Stability API with TensorRT and AMD-optimized versions.
- Modalities
- text → image
- Parameters
- 2B
- License
- Stability AI Community License
- At launch
- Open weights, API
- 2B parameters
- TensorRT-optimized version developed with NVIDIA
- Available on the Stability API powered by Fireworks AI
Apple Foundation Models
Apple introduced its first foundation models for Apple Intelligence: an on-device language model of about 3B parameters and a larger server model running on Private Cloud Compute.
- Modalities
- text → text
- Parameters
- About 3B (on-device); server size not disclosed
- At launch
- Consumer app
- Integrated into iOS 18, iPadOS 18 and macOS Sequoia
- On-device model uses low-bit palletization averaging 3.7 bits per weight
- Uses adapters; a rank-16 on-device adapter is typically tens of megabytes
Qwen2
Qwen2 is Qwen's second-generation model series of base and instruction-tuned models in five sizes, including the Qwen2-57B-A14B mixture-of-experts model. Base models were pre-trained with 32K context.
- Modalities
- text → text
- Context
- 125K context
- Parameters
- 0.5B, 1.5B, 7B, 57B-A14B, 72B
- License
- Apache-2.0 (72B uses the Tongyi Qianwen License Agreement)
- At launch
- Open weights
- Qwen2-7B-Instruct and Qwen2-72B-Instruct handle up to 128K tokens with YaRN
- Includes Qwen2-0.5B, 1.5B, 7B, 57B-A14B, and 72B
- Smaller sizes under Apache 2.0
Kling
Kling is Kuaishou's first video generation model, a diffusion transformer with a self-developed 3D VAE that generates videos of up to two minutes at 30fps and up to 1080p from text prompts.
- Modalities
- text → video
- At launch
- Consumer app
- Kuaishou's July 25, 2024 release dates the opening of trial applications to June 6, 2024
- Diffusion transformer with a self-developed 3D VAE for spatiotemporal compression
- Initial beta testing ran inside the KuaiYing app; web access and a global beta followed in July 2024
GLM-4-9B
GLM-4-9B is the open-weights series from Z.ai's GLM-4 generation, including base, chat, and long-context chat models. GLM-4-9B-Chat supports web browsing, code execution, custom tool calls, and long-text reasoning.
- Modalities
- text → text
- Context
- 125K context
- Parameters
- 9B
- License
- GLM-4 License
- At launch
- Open weights
- GLM-4-9B-Chat supports up to 128K context
- GLM-4-9B-Chat-1M supports a 1M-token context
- Adds support for 26 languages, including Japanese, Korean, and German
GLM-4V-9B
GLM-4V-9B is Z.ai's open-weights vision-language model from the GLM-4-9B series, supporting bilingual Chinese and English multi-turn dialogue on 1120 by 1120 images. It was released the same day as GLM-4-9B.
- Modalities
- text, image → text
- Context
- 8K context
- Parameters
- 9B
- License
- GLM-4 License
- At launch
- Open weights
- High-resolution 1120 by 1120 image input
- 8K context length
- Weights under the GLM-4 License
Stable Audio Open
Open text-to-audio model for generating up to 47 seconds of samples and sound effects such as drum beats, instrument riffs, ambience and foley for music production and sound design.
- Modalities
- text → audio
- Parameters
- 1.1B
- License
- Stability AI Community License
- At launch
- Open weights
- Up to 47 seconds of audio per generation
- Weights on Hugging Face
- 1.1B parameters, per Stability AI's later Stable Audio Open Small announcement
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 2024." The AI Engineering, updated Jun 28, 2024. https://theaiengineering.com/releases/2024/06
HTML link
<a href="https://theaiengineering.com/releases/2024/06">AI models released in June 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